# Agent Casefile — full casefile

> Agent Casefile tracks what AI agents actually do in real businesses, by industry and role. Every case is sourced, dated, and graded A–D on evidence strength. Deepest coverage: healthcare and MedTech commercial teams.
> License: CC BY 4.0. Attribute "Agent Casefile (agentcasefile.com)". Canonical: https://agentcasefile.com/

---

# How is Citi using the Devin agentic coding tool?

> About 14,000 Citi engineers use an agentic coding tool (Devin), per The National CIO Review. Citi reported technology vendor spending down 30% year over year and a 30%-40% improvement in product launch speed as more development moved in-house.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Citigroup (bank)
- **Industry:** Banking & Financial Services — Global bank
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- engineers using the agent: approximately 14,000 [1]
- technology vendor spending: declined 30% year over year [1]
- product launch speed: 30% – 40% improvement [1]

## How it works
1. Engineers delegate development tasks to the Devin agent.
2. More work done internally instead of by external vendors.

## Risks & compliance
- Code quality and security review of agent output
- Attribution: vendor-spend and launch-speed changes have multiple causes

## Notes
- Outcomes are company-reported in earnings commentary and relayed by an independent publication.

## Sources
1. The National CIO Review, "How Citi and Bank of America Are Measuring AI's Payoff" (2026-10-14) <https://nationalcioreview.com/articles-insights/cio-field-notes/how-citi-and-bank-of-america-are-measuring-ais-payoff/>
   > "Approximately 14,000 Citi engineers are using an agentic coding tool, helping the bank launch products faster and bring more development work in-house."

Canonical: https://agentcasefile.com/cases/citi-devin-agentic-coding-14000-engineers/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is UnitedHealth Group using agentic AI in claims and clinical workflows?

> UnitedHealth Group is expanding agentic AI through Optum Insight. Per The National CIO Review, it helps a Medicaid plan adjudicate difficult claims in seconds instead of days, chart summarization (available to more than 40% of employed clinicians) cut chart review time by 33%, and AI-assisted research shortened appeals cycles by 26%.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** UnitedHealth Group (health insurer and care delivery company)
- **Industry:** Hospitals & Health Systems — Payer & care delivery (Optum)
- **Roles:** Revenue Cycle & Prior Auth, Clinicians & Clinical Ops
- **Tools:** Optum Insight agentic AI (Optum)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chart review time: cut by 33% [1]
- appeals processing cycles: shortened by 26% [1]
- clinician access to chart summarization: more than 40% of employed clinicians [1]

## How it works
1. Agentic AI adjudicates difficult claims for a Medicaid plan client.
2. AI chart summarization for employed clinicians.
3. AI-assisted research for appeals.

## Risks & compliance
- Automated claims and appeals decisions face regulatory and public scrutiny
- Company-reported figures from earnings commentary

## MedTech angle
Payer-side automation of claims, appeals, and prior authorization affects reimbursement timelines for device-dependent procedures.

## Sources
1. The National CIO Review, "UnitedHealth Group Expands Agentic AI Across Healthcare Operations" (2026-10-13) <https://nationalcioreview.com/articles-insights/cio-field-notes/unitedhealth-group-expands-agentic-ai-across-healthcare-operations/>
   > "AI-enabled chart summarization is now available to more than 40% of UnitedHealth’s employed clinicians, cutting chart review time by 33%"

Canonical: https://agentcasefile.com/cases/unitedhealth-optum-agentic-ai-claims-chart-review/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is CarMax using AI voice agents for customer calls?

> CarMax said its AI voice technology now covers 100% of inbound calls to stores and its customer experience center (The National CIO Review). Sierra, its agent vendor, describes agents answering questions like store hours and vehicle availability and handing off to associates.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** CarMax (used-car retailer)
- **Industry:** Consumer Products & Retail — Automotive retail
- **Roles:** Customer Support, Sales Development (SDR)
- **Tools:** AI voice agents (Sierra)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- inbound call coverage: 100% of inbound calls to stores and customer experience center [1]

## How it works
1. Voice agents answer inbound store and customer-center calls.
2. Agents ask clarifying questions, answer common queries, and hand off to CarMax associates.

## Risks & compliance
- Coverage is not resolution; no resolution rate disclosed

## Sources
1. The National CIO Review, "CarMax Drives Agentic AI Across Its Customer Operations" (2026-09-30) <https://nationalcioreview.com/articles-insights/cio-field-notes/carmax-drives-agentic-ai-across-its-customer-operations/>
   > "CarMax said its AI voice technology now covers 100% of inbound calls to stores and its customer experience center."
2. Sierra, "CarMax and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/carmax>
   > "resulting in quick answers to common questions, such as store hours and vehicle availability, and smoother transitions to a CarMax associate."

Canonical: https://agentcasefile.com/cases/carmax-agentic-voice-ai-inbound-calls/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Fisher & Paykel raise self-service resolution with an AI agent?

> Fisher & Paykel, the premium home-appliance brand, replaced rule-based Einstein bots with an Agentforce service agent that troubleshoots by appliance model and books service. It told Chief Marketer the agent resolves about 70% of troubleshooting queries vs about 30% before; Salesforce later reported customers solve their own issues 80% of the time.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Fisher & Paykel (appliances) (consumer appliance maker)
- **Industry:** Consumer Products & Retail — Home appliances
- **Roles:** Customer Support, Field Service & Case Coverage
- **Tools:** Agentforce (Salesforce); Data 360 (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- self-service resolution: about 70% of product troubleshooting queries vs about 30% with prior bots [1] (self-reported)
- self-service resolution (later report): customers solve their own issues 80% of the time [2] (self-reported)

## How it works
1. Website chat agent identifies the exact appliance model.
2. Searches knowledge articles, service manuals, and CRM records (RAG).
3. Guides the customer through troubleshooting or helps book/reschedule a technician.
4. Escalates to a human with a conversation summary; a separate agent briefs technicians before visits.

## Risks & compliance
- Identity verification before appointment changes (described in source)

## Notes
- Not to be confused with Fisher & Paykel Healthcare, a separate medical device company.
- Resolution figures differ by source and date (about 70% in Dec 2025; 80% in Sep 2026).

## Sources
1. Chief Marketer, "Luxury Appliance Brand Fisher & Paykel Taps Agentic AI for Efficiency" (2025-12-01) <https://www.chiefmarketer.com/luxury-appliance-brand-fisher-paykel-taps-agentic-ai-for-efficiency/>
   > "In terms of self-service, it’s resolving about 70% of product troubleshooting queries,"
2. Salesforce, "Fisher & Paykel resolves more appliance issues with Agentforce" (2026-09-15) <https://www.salesforce.com/customer-stories/fisher-and-paykel/agentic-self-service/>
   > "Customers can now solve their own issues 80% of the time with help from Agentforce."

Canonical: https://agentcasefile.com/cases/fisher-paykel-appliances-agentforce-self-service/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How much of Airbnb's customer support is handled by AI agents?

> Airbnb says its custom-built AI agent handles roughly a third of customer support issues in the U.S. and Canada (about 30% of North American tickets, per CX Today) without a live specialist, and it plans a global rollout in all languages where it employs human agents.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Airbnb (online travel marketplace)
- **Industry:** Software, Internet & Media — Travel marketplace
- **Roles:** Customer Support
- **Tools:** Custom-built AI support agent (Airbnb (in-house))
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- share of North American support tickets handled by AI: 30% [2]

## How it works
1. Custom AI agent trained on millions of customer interactions.
2. Resolves tickets without a live specialist; escalates the rest.
3. Planned expansion to voice and chat globally.

## Risks & compliance
- Disputes, safety, and refunds need human judgment
- Company-reported figure from an earnings call

## Notes
- The 30% global figure is a one-year goal, not an outcome.

## Sources
1. TechCrunch, "Airbnb says a third of its customer support is now handled by AI in the US and Canada" (2026-02-13) <https://techcrunch.com/2026/02/13/airbnb-says-a-third-of-its-customer-support-is-now-handled-by-ai-in-the-u-s-and-canada/>
   > "Airbnb says its custom-built AI agent is now handling roughly a third of its customer support issues in North America"
2. CX Today, "AI Agents Handle 30% of Airbnb Customer Support Tickets as Company Expands Automation" (undated; accessed 2026-10-17) <https://www.cxtoday.com/ai-automation-in-cx/ai-agents-handle-30-of-airbnb-customer-support-tickets-as-company-expands-automation/>
   > "Airbnb says AI agents now handle 30% of its customer service tickets in North America as the company scales automation."

Canonical: https://agentcasefile.com/cases/airbnb-custom-ai-agent-customer-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Does ServiceNow use its own AI agents for customer, IT, and HR service?

> ServiceNow CEO Bill McDermott said AI agents now handle 90% of the company's customer service, IT service, and HR service, according to CX Foundation, which notes the company had not made layoffs.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** ServiceNow (enterprise software company)
- **Industry:** Software, Internet & Media — Enterprise SaaS
- **Roles:** Customer Support
- **Tools:** ServiceNow AI Agents (ServiceNow)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- share of service handled by AI agents: 90% (CEO statement) [1]

## How it works
1. ServiceNow runs its own AI agents across customer, IT, and HR service.
2. Humans handle the remainder.

## Risks & compliance
- CEO claim; methodology for "handled" not disclosed
- Vendor dogfooding its own product

## Notes
- Executive claim relayed by trade press; treat as self-reported despite the independent publisher.

## Sources
1. CX Foundation, "ServiceNow Claims AI Handles 90% of Its Customer Service Queries, But Hasn’t Made Layoffs" (2026-01-20) <https://cxfoundation.com/news/servicenow-claims-ai-handles-90-percent-customer-service>
   > "Bill McDermott, CEO of ServiceNow, has stated that AI agents now handle 90% of the company’s customer service queries."

Canonical: https://agentcasefile.com/cases/servicenow-ai-agents-internal-service/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Did an AI voice agent increase colorectal cancer screening sign-ups?

> WellSpan Health used 'Ana', a multilingual AI care agent from Hippocratic AI, to phone 1,878 patients due for colorectal cancer screening in September 2024. A peer-reviewed JMIR study found FIT-kit opt-in of 18.2% among Spanish-speaking patients vs 7.1% among English-speaking patients; the study was funded by Hippocratic AI.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** WellSpan Health (health system)
- **Industry:** Hospitals & Health Systems — Integrated health system
- **Roles:** Patient Access & Outreach, Clinicians & Clinical Ops
- **Tools:** Hippocratic AI care agent ('Ana') (Hippocratic AI)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- FIT test opt-in (Spanish vs English speakers): 18.2% vs 7.1% [1]
- connect rate (Spanish vs English speakers): 69.6% vs 53.0% [1]

## How it works
1. Eligible patients without active web-based health profiles are selected for outreach.
2. The AI agent calls each patient in their preferred language.
3. It explains colorectal cancer screening and answers questions.
4. Patients who opt in get a FIT kit mailed to their home.

## Risks & compliance
- Patient-facing clinical conversation; escalation path required
- Study outcome is opt-in, not completed screening

## Notes
- JMIR conflict-of-interest statement: most authors are Hippocratic AI employees and Hippocratic AI funded the study; WellSpan provided data.
- The vendor page reports an 89% Spanish-speaker connect rate; the peer-reviewed paper reports 69.6%. We use the paper's figure.
- Paper text verified via Europe PMC (PMC12278996) because jmir.org blocks automated fetches.

## Sources
1. Journal of Medical Internet Research, "Using a Multilingual AI Care Agent to Reduce Disparities in Colorectal Cancer Screening for Higher Fecal Immunochemical Test Adoption Among Spanish-Speaking Patients: Retrospective Analysis" (2025-06-25) <https://www.jmir.org/2025/1/e71211>
   > "Spanish-speaking patients demonstrated significantly higher engagement across all measures than English-speaking patients with respect to FIT test opt-in rates (18.2% vs 7.1%, P<.001)"
2. Hippocratic AI, "Colorectal Cancer Screening AI Case Study" (2024-10-24) <https://hippocraticai.com/colorectal-cancer-case-study/>
   > "Our AI agent, named Ana, is designed to conduct targeted outreach to at-risk patients, with a particular focus on Spanish-speaking populations."

Canonical: https://agentcasefile.com/cases/wellspan-hippocratic-ana-colorectal-screening-outreach/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Seattle Children's using an AI agent for clinical pathways?

> Seattle Children's built 'Pathway Assistant' with Google, trained only on its internal clinical pathway documents, to guide clinicians through diagnosis and treatment workflows for 70+ pediatric conditions. Its CMIO told HealthSystemCIO the assistant achieves more than 98% accuracy after tuning and is being rolled out in emergency and urgent care.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Seattle Children's (pediatric health system)
- **Industry:** Hospitals & Health Systems — Pediatric hospital
- **Roles:** Clinicians & Clinical Ops
- **Tools:** Gemini / Vertex AI (Google Cloud)
- **Status:** pilot · **Agent pattern:** copilot · **Autonomy:** 0/4
- **Last verified:** 2026-10-17

## Reported outcomes
- answer accuracy: more than 98% after repeated tuning and validation [1] (self-reported)

## How it works
1. Trained exclusively on Seattle Children's internal pathway documents.
2. Engages clinicians in multi-step dialogues that adapt to patient details.
3. Standalone by design: no EHR connection and no PHI.

## Risks & compliance
- Clinical decision support only; not medical advice
- Accuracy is self-reported by the organization

## Notes
- Accuracy figure is attributed to Seattle Children's CMIO in the article.

## Sources
1. HealthSystemCIO, "Seattle Children's Builds AI Assistant & Governance Model to Advance Pediatric Care" (2025-05-14) <https://healthsystemcio.com/2025/05/14/seattle-childrens-builds-ai-assistant-governance-model-to-advance-pediatric-care/>
   > "According to Lin, the assistant achieves more than 98% accuracy after repeated tuning and validation, and is being rolled out in emergency and urgent care settings."

Canonical: https://agentcasefile.com/cases/seattle-childrens-pathway-assistant/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Did Wiley's AI service agent beat its old chatbot?

> Publisher Wiley piloted Salesforce Agentforce during back-to-school peak to resolve account-access and password issues and triage registration and payment questions. Wiley told diginomica it saw a more than 40% increase in case resolution compared with its old chatbot.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Wiley (publisher)
- **Industry:** Software, Internet & Media — Publishing & education
- **Roles:** Customer Support
- **Tools:** Agentforce (Salesforce)
- **Status:** pilot · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- case resolution: more than 40% increase vs previous chatbot [1] (self-reported)

## How it works
1. Agent answers from Wiley's knowledge base.
2. Automatically resolves account access and password issues.
3. Triages registration and payment requests to the right human team.
4. Summarizes conversations for human agents.

## Risks & compliance
- Account security on password flows

## Notes
- Salesforce also cites a 213% ROI for Wiley, but that figure is for its Service Cloud implementation overall, not the agent.

## Sources
1. diginomica, "Dreamforce 24 - Wiley turns the page on productivity gains with Agentforce" (2024-09-17) <https://diginomica.com/dreamforce-24-wiley-turns-page-productivity-gains-agentforce>
   > "we’ve seen a more than 40% increase in case resolution, outperforming our old bot."

Canonical: https://agentcasefile.com/cases/wiley-agentforce-case-resolution/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is AI really answering customer emails at Octopus Energy?

> Octopus Energy CEO Greg Jackson said AI was doing the work of 250 people at the UK energy supplier, and that AI-written customer emails delivered 80% customer satisfaction versus 65% for skilled, trained staff, as reported by Business Insider in May 2023.

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Octopus Energy (energy supplier)
- **Industry:** Telecom & Energy Utilities — Energy supplier
- **Roles:** Customer Support
- **Tools:** Generative AI in company systems (Octopus Energy (in-house))
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- equivalent workload: work of 250 people [1] (self-reported)
- CSAT on AI-written emails: 80% vs 65% for trained staff [1] (self-reported)

## How it works
1. Generative AI was incorporated into company systems to write customer emails.
2. Human staff continue to handle remaining correspondence.

## Risks & compliance
- Figures come from the CEO's own statements, relayed by the publication

## Notes
- Business Insider attributes the figures to Jackson's column in The Times of London.

## Sources
1. Business Insider, "AI is doing the work of 250 people at an energy company and satisfying customers better than trained workers, CEO says" (2023-05-09) <https://www.businessinsider.com/ai-work-automation-octopus-energy-greg-jackson-2023-5>
   > "Emails written by AI delivered 80% customer satisfaction — comfortably better than the 65% achieved by skilled, trained people."

Canonical: https://agentcasefile.com/cases/octopus-energy-ai-customer-emails/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Currax Pharmaceuticals using AI agents to support pharma sales reps?

> Currax Pharmaceuticals rolled out SuperAgents, built on the ACTO life-sciences sales platform, to help reps retrieve approved clinical references and policies during HCP visits. Currax reports a 99% reduction in time-to-answer, a sixfold increase in daily usage per rep, and about 75% of its sales force using it consistently (MM+M).

- **Evidence grade:** A — Named organization, quantified outcome, independent source (news, paper, filing, or customer quote outside vendor marketing).
- **Organization:** Currax Pharmaceuticals (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Specialty pharma (obesity, migraine, sleep)
- **Roles:** Field Sales Reps
- **Tools:** SuperAgents (ACTO)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- time-to-answer: 99% reduction [1]
- adoption: about 75% of sales force using consistently; sixfold increase in daily usage per rep [1]

## How it works
1. Reps ask questions conversationally; the tool retrieves approved clinical reference documents, policy documents, and operational guides.
2. Supports sales practice and HCP strategy discussions.
3. Guardrails follow medical and legal team rules on what can be said about products.

## Risks & compliance
- Promotional compliance: answers must stay within approved, on-label materials
- Company-reported figures relayed by trade press

## MedTech angle
Direct analog for device reps: instant, compliant answers to unexpected clinician questions in the field.

## Notes
- Figures are company-reported to MM+M; graded A because the source is independent trade press, per rubric.

## Sources
1. MM+M, "Currax Pharmaceuticals is using AI to deploy ‘super agents’ in sales" (undated; accessed 2026-10-17) <https://www.mmm-online.com/news/currax-pharmaceuticals-ai-super-agents-sales/>
   > "This results in expediency, Dreher says, with Currax clocking a 99% reduction in time-to-answer for reps."

Canonical: https://agentcasefile.com/cases/currax-pharmaceuticals-acto-superagents-field-reps/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can an AI agent cut prior authorization handling time?

> MIT Health, MIT's campus health center, embedded Notable's Sidekick assistive AI agent into prior-authorization submissions to the Carelon portal; it pre-assembles clinical packets and answers payer questions from the chart. Notable reports authorizations now average about 6 minutes, an estimated 45% reduction from ~11 minutes, across 750+ runs.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** MIT Health (campus health center)
- **Industry:** Hospitals & Health Systems — Campus health center
- **Roles:** Revenue Cycle & Prior Auth
- **Tools:** Sidekick AI Co-pilot (Notable)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- prior auth handling time: ~6 minutes vs ~11 minutes (estimated 45% reduction) [1] (self-reported)
- volume: flow has run more than 750 times since go-live [1] (self-reported)

## How it works
1. Puts relevant orders and chart context in one place.
2. Pre-assembles the clinical packet for the payer.
3. Answers payer clinical questions using context from the patient's chart.
4. Errors (e.g., missing allergies) create human tasks with clear instructions.

## Risks & compliance
- PHI
- Clinical answers to payer questions should be reviewed by staff

## Notes
- Figure applies 'when all information is available', per the source.

## Sources
1. Notable, "MIT Health cut prior auth time in half with AI" (2026-06-09) <https://www.notablehealth.com/customer-stories/mit-health-cut-prior-auth-time-in-half-with-ai>
   > "A Sidekick-supported authorization can be completed in an average of 6 minutes, an estimated 45% reduction from the previous average ~11-minute handling time, when all information is available."

Canonical: https://agentcasefile.com/cases/mit-health-notable-prior-authorization/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can an AI agent build complex sales quotes in a life-sciences company?

> BioIVT, which supplies biological specimens for pharma and diagnostic research, built 'Ivy', an Agentforce quoting agent that reads an opportunity, matches requests to 15,000+ product codes, builds the quote, and routes it to a service rep for review. BioIVT says quoting went from days to minutes, one rep gained 43% more time for high-value work, and Ivy has reached 94% accuracy.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** BioIVT (life-science research supplier)
- **Industry:** Pharma, Biotech & Life Sciences — Research biospecimens supplier
- **Roles:** Sales Ops & Quoting, Customer Support
- **Tools:** Agentforce (Salesforce); Data 360 (Salesforce)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- quote turnaround: from days to minutes [1] (self-reported)
- rep capacity: one service rep has 43% more time for high-value tasks [2] (self-reported)
- agent accuracy: 94% accuracy rate [2] (self-reported)

## How it works
1. Standard catalog quote requests are assigned to Ivy as cases; a one-minute pause allows a human to pull a case first.
2. Ivy reads the opportunity (products, quantities, shipping) and matches against 15,000+ product codes.
3. It creates the quote; an outbound message fetches customer-specific pricing from the ERP.
4. A service rep reviews before sending; rep edits are captured to refine Ivy.
5. Guardrail: Ivy cannot suggest alternative products and escalates when unsure.

## Risks & compliance
- Wrong product match can change scientific outcomes (BioIVT's stated reason for its no-substitution guardrail)
- Human review before every send

## MedTech angle
Directly transferable to device quoting (kits, implant sets, capital equipment) where catalog complexity is high.

## Notes
- The 43% figure is one service rep's account, as stated in the source.
- Went live July 2025; built with partner CREO Consulting.

## Sources
1. Salesforce, "How BioIVT built agentic research quoting that gives service reps 43% more time for high-impact work" (2026-05-07) <https://www.salesforce.com/customer-stories/bioivt-larae-bennett-interview/>
   > "With Agentforce, we went from days to minutes."
2. Salesforce, "How BioIVT built agentic research quoting (same interview)" (2026-05-07) <https://www.salesforce.com/customer-stories/bioivt-larae-bennett-interview/>
   > "She has 43% more time for the high-value tasks that really require her expertise, which is exactly what we hoped to accomplish with Ivy."

Canonical: https://agentcasefile.com/cases/bioivt-ivy-quoting-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How many customer queries does Air India's AI agent handle?

> Air India's AI.g agent, built on Azure OpenAI, answers customer questions such as baggage and pet policies. Microsoft reports it has resolved more than 13 million conversations with a 97% success rate, with only 3% of queries escalated to a human agent.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Air India (airline)
- **Industry:** Transportation & Logistics — Airline
- **Roles:** Customer Support
- **Tools:** Azure OpenAI (Foundry Models) (Microsoft)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- conversations resolved: more than 13 million, 97% success rate [1] (self-reported)
- escalation: only 3% of queries escalated to a human [1] (self-reported)

## How it works
1. Customers ask AI.g questions in natural language.
2. The agent reasons over Air India policy content to answer (e.g., recognizing a Labrador falls under the pet policy).
3. Complex or unresolved queries escalate to human contact-center staff.
4. Air India is trialing agents for full workflows such as refund processing (per source).

## Risks & compliance
- Consumer-protection rules on refunds and compensation

## Sources
1. Microsoft, "How Azure AI helped Air India reinvent customer service by answering 40,000 daily queries instantly" (2026-02-11) <https://www.microsoft.com/en/customers/story/26047-air-india-azure-openai-in-foundry-models>
   > "Since launch, it has resolved more than 13 million conversations with a 97% success rate."

Canonical: https://agentcasefile.com/cases/air-india-aig-virtual-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can an AI copilot help clinicians plan cancer workups?

> Color Health built an LLM-powered 'Cancer Copilot' to assist with the workup of patients newly diagnosed with cancer. Color reports more than 95% concordance with guidelines on a dataset of anonymized real-world patients at UCSF Health.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Color Health (healthcare provider / health tech)
- **Industry:** Hospitals & Health Systems — Virtual cancer care provider
- **Roles:** Clinicians & Clinical Ops
- **Tools:** Cancer Copilot (Color Health)
- **Status:** pilot · **Agent pattern:** copilot · **Autonomy:** 0/4
- **Last verified:** 2026-10-17

## Reported outcomes
- guideline concordance: > 95% on anonymized real-world UCSF Health patients [1] (self-reported)

## How it works
1. Extracts explicit decision factors from the patient record.
2. Applies declarative guideline logic to recommend workup steps.
3. Produces an explanatory summary with a reasoning chain and guideline references, designed for expert review.

## Risks & compliance
- Clinical decision support; not medical advice
- Evaluation was on a retrospective dataset

## Sources
1. Color Health, "Expertise Is What We Want" (2026-01-26) <https://www.color.com/blog/expertise-is-what-we-want/>
   > "The system achieved a high degree of concordance with guidelines (> 95%) on a data set composed of anonymized real-world patients at UCSF Health."

Canonical: https://agentcasefile.com/cases/color-health-cancer-copilot-workup/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Medtronic using AI agents to prep device reps for sales calls?

> Medtronic's EurAsia medical education and sales-enablement team ran a Data Cloud + Agentforce proof of concept that puts training and education insights in front of field reps, with agent use cases for sales-prep summaries, T&E history/recommendations, and instant follow-ups. Salesforce reports the team delivered value in 5 weeks, with field adoption signals of 84% rating it 4+ and 90% recommending it.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Medtronic (medical device manufacturer)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Medical devices (EurAsia region)
- **Roles:** Field Sales Reps, Commercial Leadership
- **Tools:** Agentforce (Salesforce); Data Cloud (Salesforce)
- **Status:** pilot · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- time to value: value delivered in 5 weeks [1] (self-reported)
- field adoption: 84% rated 4+; 90% would recommend; 100% say Salesforce value increased [1] (self-reported)

## How it works
1. Unifies training & education (T&E) and customer touchpoint data in Data Cloud.
2. Agent produces sales-prep summaries before a rep's call.
3. Surfaces each clinician's T&E history and recommendations.
4. Drafts instant follow-ups after interactions.

## Risks & compliance
- HCP data privacy and consent
- Promotional-content compliance for follow-ups
- Identity resolution across data sources (named by Medtronic as a scaling requirement)

## MedTech angle
Directly applicable to device reps who need clinician education history before a case or meeting; a 90-day pilot plan is described as the scaling path.

## Notes
- Proof of concept / pilot stage per the source; adoption metrics are survey signals, not revenue outcomes.

## Sources
1. Salesforce, "Customer 360: Turning customer data into commercial insights in MedTech" (2025-12-04) <https://www.salesforce.com/eu/blog/turning-customer-data-into-commercial-insights-in-medtech/>
   > "How the team delivered value in 5 weeks (field adoption signals: 84% rating 4+, 90% recommend, 100% say Salesforce value increased)."

Canonical: https://agentcasefile.com/cases/medtronic-agentforce-field-sales-prep/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How many interactions has Bank of America's Erica virtual assistant handled?

> Bank of America's AI-driven virtual financial assistant Erica has surpassed 3 billion client interactions, assisting nearly 50 million users since launch and averaging more than 58 million interactions a month. BofA also says Erica for Employees has reduced calls into its IT service desk by 50%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Bank of America (bank)
- **Industry:** Banking & Financial Services — Retail banking
- **Roles:** Customer Support
- **Tools:** Erica (Bank of America (in-house))
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- client interactions: 3 billion+ total; 58 million+ per month [1] (self-reported)
- IT service desk calls (Erica for Employees): reduced by 50% [1] (self-reported)

## How it works
1. Clients ask Erica questions in the mobile app.
2. Erica delivers proactive, personalized insights (e.g., balance trends over the next 7 days).
3. An employee version handles internal IT questions.

## Risks & compliance
- Banking regulation and complaint handling

## Notes
- Erica launched in 2018 and predates LLM agents; included as a long-running virtual-assistant benchmark.

## Sources
1. Bank of America, "A Decade of AI Innovation: BofA's Virtual Assistant Erica Surpasses 3 Billion Client Interactions" (2025-08-20) <https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation--bofa-s-virtual-assistant-erica-surpas.html>
   > "surpassing 3 billion client interactions, and now averaging more than 58 million interactions per month"

Canonical: https://agentcasefile.com/cases/bank-of-america-erica-virtual-assistant/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How many shipping tasks have C.H. Robinson's AI agents performed?

> Logistics provider C.H. Robinson says its fleet of generative AI agents has performed over 3 million shipping tasks, including well over 1 million price quotes and 1 million orders processed. Its orders agent gets 5,200+ customers' loads accepted in under 90 seconds instead of waiting up to four hours in an email queue.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** C.H. Robinson (logistics provider)
- **Industry:** Transportation & Logistics — Freight brokerage / 3PL
- **Roles:** Logistics Operations, Sales Ops & Quoting
- **Tools:** Proprietary generative AI agents (C.H. Robinson)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- price quotes: well over 1 million delivered by AI [1] (self-reported)
- order acceptance time: under 90 seconds vs up to four hours [1] (self-reported)

## How it works
1. Quoting agent returns customer-specific price quotes, including LTL.
2. Orders agent reads emailed tenders and processes orders.
3. Truck-posting agent reads carrier capacity emails and feeds the real-time capacity center.

## Risks & compliance
- Pricing errors at scale
- Customer data in email ingestion

## MedTech angle
Quote-and-order agents are a template for distributor and device order desks.

## Sources
1. C.H. Robinson, "C.H. Robinson: Artificial Intelligence Has Now Performed Over 3 Million Shipping Tasks" (2025-04-16) <https://www.chrobinson.com/en-us/about-us/newsroom/press-releases/2025/ai-performs-over-three-million-shipping-tasks/>
   > "Instead of waiting up to four hours for a person to get to their shipment in an email queue, over 5,200 customers are getting their loads accepted in under 90 seconds."

Canonical: https://agentcasefile.com/cases/ch-robinson-freight-quoting-order-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can an AI agent handle tax-season customer support surges?

> 1-800Accountant deployed Agentforce for 24/7 client support during tax season, answering questions such as tax return status using CRM data and IRS public sources. Its CTO says Agentforce now helps resolve up to 50% of incoming requests.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** 1-800Accountant (accounting firm)
- **Industry:** Professional Services — Accounting & tax
- **Roles:** Customer Support
- **Tools:** Agentforce (Salesforce); Data Cloud (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- requests resolved: up to 50% of incoming requests [1] (self-reported)

## How it works
1. Agent reasons across client data, audit logs, and support history harmonized in Data Cloud.
2. Uses trusted public sources such as the IRS website for tax questions.
3. Frees human staff for complex tasks.

## Risks & compliance
- Tax advice liability
- Client financial data security

## Sources
1. Salesforce, "1-800Accountant Adopts Salesforce's Agentforce to Scale Expertise During Tax Season Surge" (2025-03-04) <https://www.salesforce.com/news/press-releases/2025/03/04/1800accountant-agentforce-customer-support/>
   > "With Agentforce now helping to resolve up to 50% of incoming requests, we can securely address responses to questions like tax return status"

Canonical: https://agentcasefile.com/cases/1800accountant-agentforce-tax-season-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How are Morgan Stanley's financial advisors using AI assistants?

> Morgan Stanley Wealth Management launched AI @ Morgan Stanley Debrief, an OpenAI-powered tool that, with client consent, generates meeting notes, surfaces action items, drafts a follow-up email for the advisor to edit, and saves a note into Salesforce. The firm reported that 98% of Financial Advisor teams had adopted its earlier AI @ Morgan Stanley Assistant.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Morgan Stanley Wealth Management (wealth manager)
- **Industry:** Banking & Financial Services — Wealth management
- **Roles:** Financial Advisors & Bankers
- **Tools:** AI @ Morgan Stanley Debrief (Morgan Stanley (OpenAI-powered))
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- adoption (AI @ Morgan Stanley Assistant): 98% of Financial Advisor teams [1] (self-reported)

## How it works
1. With client consent, Debrief generates notes during client meetings and surfaces action items.
2. After the meeting it summarizes key points.
3. Drafts an email for the advisor to edit and send at their discretion.
4. Saves a note into Salesforce.

## Risks & compliance
- Client consent for recording
- Advisor reviews all client communications

## MedTech angle
Same post-meeting pattern (notes → CRM → follow-up draft) used by device reps after surgeon meetings.

## Notes
- The 98% adoption figure refers to the AI @ Morgan Stanley Assistant, not Debrief.

## Sources
1. Morgan Stanley, "Morgan Stanley Wealth Management Announces Latest Game-Changing Addition to Suite of GenAI Tools" (2024-06-26) <https://www.morganstanley.com/press-releases/ai-at-morgan-stanley-debrief-launch>
   > "To date, 98% of Financial Advisor teams have adopted the Assistant."

Canonical: https://agentcasefile.com/cases/morgan-stanley-ai-debrief-advisors/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How much time does a sales AI copilot save sellers at Lumen?

> Lumen Technologies uses Microsoft Copilot to cut a seller's customer-outreach research from about four hours to 15 minutes. Lumen values four hours back per seller each week at $50 million in revenue over 12 months.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Lumen Technologies (telecom carrier)
- **Industry:** Telecom & Energy Utilities — Telecommunications
- **Roles:** Field Sales Reps, Sales Ops & Quoting
- **Tools:** Microsoft 365 Copilot (Microsoft)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 0/4
- **Last verified:** 2026-10-17

## Reported outcomes
- research time per outreach: four hours to 15 minutes [1] (self-reported)
- estimated value: $50 million in revenue over 12 months (company estimate) [1] (self-reported)

## How it works
1. Seller asks Copilot to compile account research.
2. Copilot drafts a briefing used to prepare customer outreach.

## Risks & compliance
- Revenue value is an estimate, not realized revenue

## MedTech angle
Account research before a hospital or surgeon meeting is the same pattern for device reps.

## Sources
1. Microsoft, "Lumen's strategic leap: How Copilot is redefining productivity and employee engagement" (2024-05-21) <https://www.microsoft.com/en/customers/story/1771760434465986810-lumen-microsoft-copilot-telecommunications-en-united-states>
   > "It typically takes a seller four hours to do research for customer outreach, and with generative AI, they can now do that in 15 minutes."

Canonical: https://agentcasefile.com/cases/lumen-copilot-seller-research/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How much customer service work did Klarna's AI assistant take on?

> Klarna's OpenAI-powered assistant had 2.3 million conversations in its first month, two-thirds of Klarna's customer service chats, which Klarna equated to the work of 700 full-time agents. Klarna also reported errands resolved in under 2 minutes vs 11 minutes previously.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Klarna (payments company)
- **Industry:** Banking & Financial Services — Payments / consumer finance
- **Roles:** Customer Support
- **Tools:** OpenAI models (OpenAI)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- share of chats: 2.3 million conversations, two-thirds of customer service chats [1] (self-reported)
- equivalent workload: work of 700 full-time agents [1] (self-reported)
- resolution time: under 2 minutes vs 11 minutes [1] (self-reported)

## How it works
1. Available in the Klarna app 24/7 in 23 markets and 35+ languages.
2. Handles refunds, returns, payment issues, cancellations, disputes, and invoice inaccuracies.
3. Customers can still choose a live agent.

## Risks & compliance
- Financial-services complaint handling rules
- Escalation to humans must stay available

## Notes
- Figures are from the launch month (Feb 2024) as stated by Klarna.

## Sources
1. Klarna, "Klarna AI assistant handles two-thirds of customer service chats in its first month" (2024-02-27) <https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/>
   > "The AI assistant has had 2.3 million conversations, two-thirds of Klarna’s customer service chats It is doing the equivalent work of 700 full-time agents"

Canonical: https://agentcasefile.com/cases/klarna-ai-assistant-customer-service/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# What resolution rate did WeightWatchers get from its AI care agent?

> WeightWatchers built a member-care AI agent with Sierra, first tested in its mobile app on a small share of traffic, then expanded. Sierra reports the agent contained nearly 70% of all cases in its first week with a customer satisfaction score above 4.5/5.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** WeightWatchers (consumer subscription company)
- **Industry:** Consumer Products & Retail — Consumer wellness subscription
- **Roles:** Customer Support
- **Tools:** Sierra agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- containment: nearly 70% of cases in the first week [1] (self-reported)
- CSAT: greater than 4.5/5 [1] (self-reported)

## How it works
1. Agent loaded with WeightWatchers knowledge base, policies, and brand guidelines plus guardrails.
2. Piloted in the mobile app on a small percentage of traffic.
3. IT team stress-tested the agent with questions written to trip it up before broad rollout.

## Risks & compliance
- Health-adjacent conversations; avoid medical advice
- Customer data protection

## Sources
1. Sierra, "How WeightWatchers embraces AI to engage members with empathy, at scale, with Sierra" (2024-02-13) <https://sierra.ai/customers/weightwatchers>
   > "With Sierra, WeightWatchers AI agent was able to contain nearly 70% of all cases within the first week, while continuing to deliver a greater than 4.5/5 score in customer satisfaction."

Canonical: https://agentcasefile.com/cases/weightwatchers-sierra-member-care/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Zembl use AI agents to raise sales conversion?

> Zembl, an Australian business energy consultant, built a 24/7 AI agent workforce with Relevance AI across sales, marketing, and operations to handle rising inbound leads and after-hours inquiries. Zembl reports a 30% increase in overall customer conversion without adding headcount.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Zembl (energy brokerage / consultancy)
- **Industry:** Telecom & Energy Utilities — Business energy comparison (Australia)
- **Roles:** Sales Development (SDR), Sales Ops & Quoting
- **Tools:** AI agents (Relevance AI)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- customer conversion: 30% increase without adding headcount [1] (self-reported)

## How it works
1. Agents respond to inbound leads, including after hours and weekends.
2. Agents take on sales and operations admin to save rep time.

## Risks & compliance
- Outbound messaging compliance (spam, consent rules)
- Vendor-published results; not independently audited

## Sources
1. Relevance AI, "Zembl built a 24/7 AI sales workforce — and grew customer conversion by 30% without adding headcount" (undated; accessed 2026-10-17) <https://relevanceai.com/customers/zembl>
   > "we’ve seen a 30% increase in overall customer conversion."

Canonical: https://agentcasefile.com/cases/zembl-relevance-ai-sales-workforce/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is ZayZoon using an AI agent for support?

> ZayZoon, an earned wage access fintech, uses Fin for support. Intercom reports an 80% Fin resolution rate.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** ZayZoon (earned wage access fintech)
- **Industry:** Banking & Financial Services — Fintech (Canada)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 80% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Financial-services account security

## Sources
1. Intercom (Fin), "How ZayZoon saves millions with Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/zayzoon>
   > "How ZayZoon saves millions with Fin"

Canonical: https://agentcasefile.com/cases/zayzoon-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did the World Economic Forum use AI agents for Davos attendees?

> The World Economic Forum used Salesforce Agentforce to answer attendee questions on sessions, venues, and shuttles at its Annual Meeting. Salesforce reports a 99.9% response rate and 75% of attendee inquiries resolved at Davos 2025.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** World Economic Forum (international organization / event host)
- **Industry:** Professional Services — International organization & events
- **Roles:** Customer Support
- **Tools:** Agentforce (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- inquiries resolved: 75% [1] (self-reported)
- response rate: 99.9% [1] (self-reported)

## How it works
1. Agent answers session, venue, and shuttle questions across 450+ sessions.
2. Escalates the rest to staff.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Salesforce, "World Economic Forum and Agentforce" (undated; accessed 2026-10-17) <https://www.salesforce.com/customer-stories/world-economic-forum>
   > "At Davos 2025, Agentforce achieved a 99.9% response rate and resolved 75% of attendee inquiries with accurate, automated answers."

Canonical: https://agentcasefile.com/cases/world-economic-forum-agentforce-davos-attendee-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is WHOOP using an AI agent for pre-purchase sales questions?

> WHOOP uses Fin for pre-purchase questions that often waited 10+ hours for a reply. Intercom reports a 130% increase in sales attributed to the team and Fin resolving up to 84% of inbound questions it handles.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** WHOOP (wearable fitness technology company)
- **Industry:** Consumer Products & Retail — Consumer wearables
- **Roles:** Sales Development (SDR), Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- attributed sales: 130% increase [1] (self-reported)
- Fin resolution rate: up to 84% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "WHOOP and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/whoop>
   > "What followed was a transformation in pre-purchase support: a 130% increase in sales attributed to the team, a Fin-powered funnel that resolves up to 84% of inbound questions it's involved with"

Canonical: https://agentcasefile.com/cases/whoop-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Vivid Seats using AI agents for customer support?

> Vivid Seats uses a Sierra AI agent for fan support. Sierra reports containment up 40% and customer satisfaction up 35% since launch.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Vivid Seats (online ticket marketplace)
- **Industry:** Consumer Products & Retail — Ticket marketplace
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- containment: increased by 40% [1] (self-reported)
- customer satisfaction: risen by 35% [1] (self-reported)

## How it works
1. AI agent handles ticket and order questions.
2. Hands off to human agents when needed.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "How Vivid Seats is elevating customer experience for millions of fans with AI" (undated; accessed 2026-10-17) <https://sierra.ai/customers/vivid-seats>
   > "Since launch, containment has increased by 40% and customer satisfaction has risen by 35%."

Canonical: https://agentcasefile.com/cases/vivid-seats-sierra-customer-experience/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Vanta using an AI agent for customer support?

> Vanta replaced an AI tool that deflected around 45% of chats with Intercom's Fin, which resolved roughly 73% of cases in testing. Intercom reports a 71% Fin resolution rate.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Vanta (compliance automation software company)
- **Industry:** Software, Internet & Media — SaaS (security compliance)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 71% [1] (self-reported)
- prior AI deflection: around 45% of chat conversations [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Vanta and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/vanta>
   > "Fin was tested against their existing AI on resolution rate, accuracy, and answer quality, and resolved roughly 73% of cases."

Canonical: https://agentcasefile.com/cases/vanta-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Valon using voice AI agents in mortgage servicing?

> Valon, a mortgage servicer, uses Decagon voice AI on inbound calls. Decagon reports 50%+ voice deflection even while onboarding 80,000 loans in a month, CSAT around 90%, and average speed to answer under one minute.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Valon (mortgage servicing company)
- **Industry:** Banking & Financial Services — Mortgage servicing
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- voice deflection: 50%+ even when onboarding 80,000 loans in a single month [1] (self-reported)
- CSAT: ~90% [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Regulated mortgage servicing disclosures

## Sources
1. Decagon, "Valon and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/valon>
   > "Achieved 50%+ deflection on voice for inbound calls, even when onboarding 80,000 loans in a single month."

Canonical: https://agentcasefile.com/cases/valon-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Underdog handle game-day support spikes with an AI agent?

> Underdog, a sports gaming company, faced support spikes of 7,000+ conversations in 48 hours on big game weekends. Intercom reports Fin deflects nearly 80% of peak-volume queries while meeting regulatory guardrails.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Underdog (sports gaming company)
- **Industry:** Consumer Products & Retail — Fantasy sports & betting
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- peak deflection: nearly 80% of peak-volume queries [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Gaming regulation and responsible-gambling rules

## Sources
1. Intercom (Fin), "Underdog and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/underdog>
   > "Today, Fin deflects nearly 80% of peak-volume queries instantly, while still ensuring the team has real-time control during live events and regulatory guardrails are met every time."

Canonical: https://agentcasefile.com/cases/underdog-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Trendyol using n8n to run AI agents for sellers and staff?

> Trendyol runs a seller chatbot, a legal AI assistant in Slack, and a search relevancy agent on n8n. The seller chatbot routes questions through an intent-classifying orchestrator to about ten n8n workflows and is live to roughly 10% of 300,000 sellers; n8n reports 700 active production workflows.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Trendyol (e-commerce marketplace)
- **Industry:** Consumer Products & Retail — E-commerce marketplace (Türkiye)
- **Roles:** Customer Support, Software Engineering
- **Tools:** n8n (n8n)
- **Status:** pilot · **Agent pattern:** multi-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chatbot rollout: roughly 10% of 300,000 sellers [1] (self-reported)
- production workflows: 700 active workflows [1] (self-reported)

## How it works
1. Seller chatbot orchestrator classifies intent and calls n8n workflows (e.g., revenue over last seven days).
2. Legal bot uses RAG over internal legal docs; legal team validates answers.
3. Search relevancy agent scores the first 40 results for low-CTR queries in four storefronts.

## Risks & compliance
- Incorrect analytics answers to sellers
- Legal answers need human validation
- Vendor-published

## Notes
- Seller chatbot is in partial rollout, so status is recorded as pilot.

## Sources
1. n8n, "Trendyol case study" (undated; accessed 2026-10-17) <https://n8n.io/case-studies/trendyol/>
   > "The chatbot is currently live to roughly 10% of Trendyol's 300,000 sellers, with full rollout planned by the end of the year."

Canonical: https://agentcasefile.com/cases/trendyol-n8n-seller-chatbot-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did TidalHealth Orthopedics reduce abandoned calls with voice AI?

> TidalHealth's orthopedic practice uses an Assort Health voice agent to answer every inbound call and schedule, cancel, or reschedule directly. The vendor reports a 75% reduction in abandoned calls, waits cut from 90 minutes to seconds, and 4.5-4.6/5 patient satisfaction.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** TidalHealth Peninsula Orthopedics Institute (hospital-affiliated orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI agent (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- abandoned calls: 75% reduction [1] (self-reported)
- wait times: from 90 minutes to seconds [1] (self-reported)
- patient satisfaction: 4.5-4.6 out of 5 [1] (self-reported)

## How it works
1. Agent answers every inbound call.
2. Schedules, cancels, or reschedules directly, or routes to a live agent.
3. Post-call surveys measure patient sentiment.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Sources
1. Assort Health, "TidalHealth Orthopaedic Cuts Abandoned Calls by 75%" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/tidalhealth-orthopedics-customer-story>
   > "Our call wait times used to be up to an hour and a half — and now they’re down to seconds."

Canonical: https://agentcasefile.com/cases/tidalhealth-orthopedics-assort-voice-ai/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did ThedaCare automate care gap outreach?

> ThedaCare (7 hospitals, Wisconsin) used Notable to automate outreach to patients with open care gaps. Notable reports 900+ automated outreaches in 3 months and 350 hours of manual staff work saved.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** ThedaCare (health system)
- **Industry:** Hospitals & Health Systems — Regional health system
- **Roles:** Patient Access & Outreach
- **Tools:** Intelligent Population Health (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- outreach volume: 900+ automated care gap outreaches in 3 months [1] (self-reported)
- staff time saved: 350 hours [1] (self-reported)

## How it works
1. Identifies patients with open care gaps.
2. Sends automated outreach so patients can schedule.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "ThedaCare conducts 900+ automated care gap outreaches in 3 months" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/thedacare-conducts-900-automated-care-gap-outreaches-in-3-months>
   > "How a leading population health organization saved its staff 350 hours of manual work in care gap outreach."

Canonical: https://agentcasefile.com/cases/thedacare-notable-care-gap-outreach/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Teachable using an AI agent for creator and student support?

> Teachable (100,000 creators, nearly 40 million students) uses Fin for round-the-clock support. Intercom reports Fin involved in 85% of conversations and resolving 68% end to end.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Teachable (online course platform)
- **Industry:** Software, Internet & Media — Creator economy SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 68% [1] (self-reported)
- Fin involvement rate: 85% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Teachable and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/teachable>
   > "Today, Fin is involved in 85% of conversations and resolves 68% of queries it touches end to end, supporting creators and students around the clock."

Canonical: https://agentcasefile.com/cases/teachable-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Synthesia handle a 690% spike in support contacts with an AI agent?

> Synthesia's monthly support-seeking customers rose from 40,000 to 316,000 in four months. Intercom reports it absorbed the 690% increase without adding headcount, with a 55% Fin resolution rate.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Synthesia (AI video software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- support demand growth absorbed: 40,000 to 316,000 monthly (690%) without adding headcount [1] (self-reported)
- Fin resolution rate: 55% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "How Synthesia managed a 690% spike in customer contact" (undated; accessed 2026-10-17) <https://fin.ai/customers/synthesia>
   > "In just four months, the number of Synthesia’s customers seeking support on a monthly basis rose from 40,000 to 316,000 – a 690% increase."

Canonical: https://agentcasefile.com/cases/synthesia-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Swyftx using an AI agent for crypto exchange support?

> Swyftx, an Australian crypto exchange, uses Fin for support. Intercom reports a 48.5% increase in automated resolution rate and 40+ hours a week saved for the support team.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Swyftx (cryptocurrency exchange)
- **Industry:** Banking & Financial Services — Crypto exchange (Australia)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- automated resolution rate: 48.5% increase [1] (self-reported)
- team time saved: over 40 hours per week [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Financial-services account security

## Sources
1. Intercom (Fin), "Swyftx and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/swyftx>
   > "Since implementing Fin, Swyftx has seen a remarkable 48.5% increase in automated resolution rate, saving the team over 40 hours per week."

Canonical: https://agentcasefile.com/cases/swyftx-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Substack using AI agents for user support?

> Substack (35 million+ active subscriptions) uses a Decagon AI agent for user support. Decagon reports it resolves more than 90% of user questions without human intervention.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Substack (subscription publishing platform)
- **Industry:** Software, Internet & Media — Publishing platform
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution without humans: more than 90% of user questions [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "Substack and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/substack>
   > "Decagon implemented an AI Agent capable of handling and resolving more than 90% of user inquiries, significantly reducing the burden on the existing support team."

Canonical: https://agentcasefile.com/cases/substack-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Southwest General use an AI voice assistant for insurance calls?

> Southwest General (Ohio) deployed HOPE, a Notable AI voice assistant, to answer inbound insurance FAQ calls 24/7 during payer contract negotiations. Notable reports it answers 100% of those calls and saves $4 per call.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Southwest General Health Center (community health system)
- **Industry:** Hospitals & Health Systems — Community hospital
- **Roles:** Patient Access & Outreach, Customer Support
- **Tools:** HOPE AI voice assistant (Notable)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- coverage of insurance FAQ calls: 100% answered [1] (self-reported)
- cost per call: $4 saved per call [1] (self-reported)

## How it works
1. Voice AI answers inbound insurance and network-status FAQ calls.
2. Runs 24/7 with consistent messaging during payer negotiations.
3. Escalates calls needing a live agent.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Southwest General: HOPE AI Voice Cuts Hold Times" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/southwest-general-health-center-hope-ai-voice-assistant>
   > "HOPE helps patients get the information they need while easing pressure on our call center staff."

Canonical: https://agentcasefile.com/cases/southwest-general-notable-hope-voice-assistant/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did an orthopedic practice clear a 1,000-voicemail backlog with an AI agent?

> South Shore Orthopedics (Hingham, MA; 15 physicians) deployed an Assort Health voice agent named "Oscar" to answer calls 24/7. The vendor reports end-of-day voicemails falling to zero from a 600-1,000 backlog, a 100% provider schedule fill rate maintained, and a 4.4/5 patient satisfaction rating.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** South Shore Orthopedics (private orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics (private practice)
- **Roles:** Patient Access & Outreach
- **Tools:** AI patient access agent ("Oscar") (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- voicemail backlog: end-of-day voicemails down from 600-1,000 backlog to 0 [1] (self-reported)
- schedule fill rate: 100% maintained [1] (self-reported)
- patient satisfaction: 4.4/5 [1] (self-reported)

## How it works
1. Voice agent answers inbound calls and resolves scheduling requests.
2. Provides after-hours access instead of an answering service.
3. Hands off complex calls to front-office staff.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Sources
1. Assort Health, "South Shore Orthopedics Clears Its Voicemail Backlog and Keeps Schedules Full" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/south-shore-orthopedics-eliminates-its-voicemail-backlog-keeps-schedules-full-with-assort-health>
   > "For weeks at a stretch, the practice had 600 to 1,000 voicemails sitting in its box that staff simply couldn't get to."

Canonical: https://agentcasefile.com/cases/south-shore-orthopedics-assort-voicemail-backlog/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is [solidcore] using a voice AI agent for studio calls?

> [solidcore] piloted Fin on voice after hours and in a few locations. Intercom reports $569k in annual cost savings and 12.6k hours saved annually.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** [solidcore] (boutique fitness studio chain)
- **Industry:** Consumer Products & Retail — Fitness studios
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- annual cost savings: $569k [1] (self-reported)
- hours saved annually: 12.6k [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "[solidcore] and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/solidcore>
   > "It began as a quiet pilot after hours and in a few locations to see how well it could handle real-time voice conversations."

Canonical: https://agentcasefile.com/cases/solidcore-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is SoftBank using an AI agent for SMB customer inquiries?

> SoftBank's corporate business uses a Sierra AI agent for small-business customer inquiries. Sierra reports the self-resolution rate rising from around 50% at launch to 70%, on about 100 inquiries per day.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SoftBank Corp. (telecommunications company)
- **Industry:** Telecom & Energy Utilities — Corporate/SMB telecom (Japan)
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- self-resolution rate: from around 50% at launch to 70% [1] (self-reported)
- volume: approximately 100 inquiries per day [1] (self-reported)

## How it works
1. AI agent answers SMB customer inquiries across SoftBank corporate services.
2. Iteratively tuned to better understand intent.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "SoftBank and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/softbank>
   > "Currently, the AI agent handles approximately 100 inquiries per day, with 70% resulting in customers finding the information they need and resolving their inquiries on their own."

Canonical: https://agentcasefile.com/cases/softbank-sierra-smb-support-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is SoFi using AI agents for member support?

> SoFi (13.7 million members) uses a Sierra AI agent for member support. Sierra reports 61% containment across more than 50,000 conversations weekly.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SoFi (digital finance company)
- **Industry:** Banking & Financial Services — Consumer fintech / bank
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- containment: 61% [1] (self-reported)
- volume: more than 50,000 conversations weekly [1] (self-reported)

## How it works
1. AI agent answers member questions across banking, investing, and lending products.
2. Escalates to human agents when needed.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited
- Regulated financial disclosures and identity verification

## Sources
1. Sierra, "SoFi and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/sofi>
   > "The AI agent achieved 61% containment, handling more than 50,000 conversations weekly."

Canonical: https://agentcasefile.com/cases/sofi-sierra-member-support-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Singtel using AI agents in customer care?

> Singtel uses a Sierra AI agent for mobile and home troubleshooting and roaming sign-ups. Sierra reports 73% of troubleshooting cases and 76% of roaming sign-up requests completed without a Customer Care officer.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Singtel (telecommunications carrier)
- **Industry:** Telecom & Energy Utilities — Mobile and home telecom (Singapore)
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- troubleshooting resolved without human: 73% [1] (self-reported)
- roaming sign-ups completed without human: 76% [1] (self-reported)

## How it works
1. AI agent walks customers through mobile and home troubleshooting.
2. Completes roaming sign-up requests end to end.
3. Hands off to Customer Care officers when needed.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "Singtel and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/singtel>
   > "73% of mobile and home troubleshooting cases were resolved without requiring a Customer Care officer."

Canonical: https://agentcasefile.com/cases/singtel-sierra-troubleshooting-roaming-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is SimplePractice using AI agents for clinician-customer support?

> SimplePractice, practice-management software used by therapists and other clinicians, uses Decagon AI agents for support. Decagon reports 85% deflection within one month, beating a 66% end-of-year target.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SimplePractice (practice management software company (behavioral health))
- **Industry:** Software, Internet & Media — Health practice-management SaaS
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- deflection: 85% within one month (target 66%) [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "SimplePractice and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/simplepractice>
   > "Within one month of going live, SimplePractice exceeded its end-of-year deflection target of 66%, reaching 85%."

Canonical: https://agentcasefile.com/cases/simplepractice-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did SENTA Partners use AI voice agents for patient scheduling?

> SENTA Partners, an MSO for 15 ENT and allergy practices, uses Assort Health voice AI agents to answer inbound calls and schedule appointments. Assort reports $1.3M in added appointment revenue, $400K+ in labor costs avoided, and 250+ staff hours saved per month.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SENTA Partners (physician-led ENT/allergy management services organization)
- **Industry:** Hospitals & Health Systems — ENT & allergy practices (MSO)
- **Roles:** Patient Access & Outreach
- **Tools:** Patient access voice AI agents (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- appointment revenue: $1.3M additional appointment revenue captured [1] (self-reported)
- labor costs avoided: $400K+ [1] (self-reported)
- staff time: 250+ hours saved per month [1] (self-reported)
- hold times: cut by 97% [1] (self-reported)

## How it works
1. Inbound voice AI agents greet callers instead of putting them on hold.
2. Agents book and manage appointments and route complex calls to staff.
3. Outbound scheduling campaigns convert patients to booked appointments.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## Notes
- Before Assort, the vendor story says 24.3% of inbound calls were dropped.

## Sources
1. Assort Health, "SENTA Captures $1.3M in Revenue and Cuts Hold Times by 97%" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/senta-customer-story>
   > "We’re handling more calls and scheduling more appointments without adding headcount."

Canonical: https://agentcasefile.com/cases/senta-partners-assort-voice-ai-patient-access/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Send Payments using an AI agent workforce?

> Send Payments built interconnected Relevance AI agents for 24/7 customer service and compliance tasks. Relevance AI reports 40 hours a week given back to the team.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Send Payments (payment infrastructure provider)
- **Industry:** Banking & Financial Services — Payments (Australia)
- **Roles:** Customer Support, Sales Development (SDR)
- **Tools:** AI agents (Relevance AI)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- time saved: 40 hours per week [1] (self-reported)

## How it works
1. Specialized agents deliver 24/7 customer service.
2. Agents automate compliance tasks.
3. Agents are connected to automate whole processes.

## Risks & compliance
- Outbound messaging compliance (spam, consent rules)
- Vendor-published results; not independently audited
- Financial compliance tasks require human sign-off

## Sources
1. Relevance AI, "Inside Send Payments' AI-first strategy" (undated; accessed 2026-10-17) <https://relevanceai.com/customers/send-payments>
   > "create specialized agents that deliver 24/7 customer service, automate compliance tasks, and give back 40 hours per week to their team."

Canonical: https://agentcasefile.com/cases/send-payments-relevance-ai-agent-workforce/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is ScottsMiracle-Gro using an AI agent for customer questions?

> ScottsMiracle-Gro uses a Sierra AI agent for consumer questions. Sierra reports containment rising from 40% to 65% after curated content and knowledge were added.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** ScottsMiracle-Gro (consumer lawn and garden products company)
- **Industry:** Consumer Products & Retail — Consumer products
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- containment: from 40% to 65% [1] (self-reported)

## How it works
1. AI agent answers lawn and garden product questions.
2. Knowledge base curated to improve answers.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "ScottsMiracle-Gro and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/scottsmiraclegro>
   > "After adding curated content and knowledge, ScottsMiracle-Gro saw their containment rates jump from 40% to 65%."

Canonical: https://agentcasefile.com/cases/scottsmiraclegro-sierra-lawn-garden-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did SafetyCulture triple meetings booked with an AI BDR agent?

> SafetyCulture (75,000+ customers) deployed a Relevance AI agent that answers customer questions and books meetings on autopilot. Relevance AI reports a 3x increase in meetings booked, an 83% reduction in cost per meeting, and a 200% increase in qualified opportunities.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SafetyCulture (workplace operations software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Sales Development (SDR)
- **Tools:** Bosh AI BDR agent (Relevance AI)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- meetings booked: 3x increase [1] (self-reported)
- cost per meeting: 83% reduction [1] (self-reported)
- qualified opportunities: 200% increase [1] (self-reported)

## How it works
1. AI agent trained on SafetyCulture processes and knowledge base.
2. Answers inbound questions and books meetings for sales.

## Risks & compliance
- Outbound messaging compliance (spam, consent rules)
- Vendor-published results; not independently audited

## Sources
1. Relevance AI, "SafetyCulture recruited an AI Agent and 3x'd meetings" (undated; accessed 2026-10-17) <https://relevanceai.com/customers/safetyculture>
   > "83% reduction in cost per meeting"

Canonical: https://agentcasefile.com/cases/safetyculture-relevance-ai-bdr-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did SaaStr turn warm leads into sales with AI outreach agents?

> SaaStr uses Salesforce Agentforce to send personalized outreach to previously overlooked leads, follow up, and hand them to a rep when ready. Salesforce reports $2.7 million in additional revenue closed with autonomous outreach.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** SaaStr (B2B media and events company)
- **Industry:** Software, Internet & Media — B2B media & events
- **Roles:** Sales Development (SDR)
- **Tools:** Agentforce (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- revenue closed via autonomous outreach: $2.7 million [1] (self-reported)

## How it works
1. Agent engages overlooked warm leads with personalized outreach.
2. Follows up and hands off to a sales rep when the lead is ready.

## Risks & compliance
- Email consent and deliverability
- Vendor-published results

## Sources
1. Salesforce, "Agentic outreach turns SaaStr’s warm leads into $2.7 million in sales" (undated; accessed 2026-10-17) <https://www.salesforce.com/customer-stories/saastr-ai/agentic-outreach>
   > "Agentforce’s personalized outreach engages previously overlooked leads, follows up, and hands them off to a sales rep when they’re ready to buy."

Canonical: https://agentcasefile.com/cases/saastr-agentforce-agentic-outreach/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How much of Rohlik Group's code is written by AI agents?

> Rohlik Group, an online grocer in five countries, describes its engineering organization as "agent-mostly." Cognition reports about 90% of code is agent-generated, with agents in planning, code generation, code review, and incident response.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Rohlik Group (online grocery company)
- **Industry:** Consumer Products & Retail — Online grocery (Europe)
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- agent-generated code: about 90% [1] (self-reported)

## How it works
1. Agents operate in every SDLC step: planning, coding, review, incident response.
2. Engineers supervise agents.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "Rohlik Group and Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/rohlikgroup>
   > "About 90% of code is agent-generated, and agents operate in every step of the SDLC: planning, code generation, code review, and incident response."

Canonical: https://agentcasefile.com/cases/rohlik-group-agent-mostly-engineering/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Rocket Mortgage using AI agents for chat and outbound calls?

> Rocket Mortgage grew a Sierra AI agent from a proof of concept to more than 400,000 successful chat conversations and over one million outbound dials each month, per Sierra.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Rocket Mortgage (mortgage lender)
- **Industry:** Banking & Financial Services — Mortgage lending
- **Roles:** Customer Support, Sales Development (SDR)
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- monthly volume: more than 400,000 successful chat conversations and over one million outbound dials each month [1] (self-reported)

## How it works
1. AI agent handles client chat conversations.
2. Runs outbound dialing at scale.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited
- Regulated financial disclosures and identity verification

## Sources
1. Sierra, "Rocket Mortgage and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/rocket-mortgage>
   > "What started as a proof of concept in May has grown to more than 400,000 successful chat conversations and over one million outbound dials each month"

Canonical: https://agentcasefile.com/cases/rocket-mortgage-sierra-chat-outbound-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Rocket Money using an AI agent for support?

> Rocket Money uses Fin for member support. Intercom reports a 68% Fin resolution rate, 54% involvement, and about $1M in annual ROI.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Rocket Money (personal finance app)
- **Industry:** Banking & Financial Services — Consumer fintech
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 68% [1] (self-reported)
- annual ROI: ~$1M [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Financial-services disclosures and account security

## Sources
1. Intercom (Fin), "Rocket Money and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/rocket-money>
   > "Rocket Money is a leading personal finance app that supports millions of users on their financial journeys"

Canonical: https://agentcasefile.com/cases/rocket-money-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Rippling using AI agents for customer support?

> Rippling added Decagon AI agents to chat and email. Decagon reports chat deflection rising from 38% to over 50% and a 7% improvement in conversation routing.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Rippling (HR and payroll software company)
- **Industry:** Software, Internet & Media — HR SaaS
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chat deflection: from 38% to over 50% [1] (self-reported)
- conversation routing: 7% improvement [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "Rippling and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/rippling>
   > "This increased chat deflection from 38% to over 50%."

Canonical: https://agentcasefile.com/cases/rippling-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Ripple Treasury use Devin for bug triage?

> Ripple's treasury platform team used Devin to investigate the codebase and work its bug backlog. Cognition reports the untriaged backlog fell 75% and bug fixes rose from roughly 0.5 to 1.5 per day, a 3x increase.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Ripple (payments and treasury software company)
- **Industry:** Banking & Financial Services — Treasury management software
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- untriaged backlog: fell by 75% [1] (self-reported)
- bugs fixed per day: from roughly 0.5 to 1.5 (3x) [1] (self-reported)

## How it works
1. Devin answers codebase questions that previously took hours or days.
2. Triages and fixes bugs; senior engineers review.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "Ripple Treasury and Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/ripple>
   > "The untriaged backlog fell by 75%."

Canonical: https://agentcasefile.com/cases/ripple-treasury-devin-bug-triage/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Reid Health automate pre-visit registration for Medicare patients?

> Reid Health (Indiana) deployed Notable automated pre-visit registration, including the Medicare Secondary Payer Questionnaire. Notable reports 53% of Medicare patients completing the MSPQ digitally, 56% engagement among geriatric patients, and 1600 staff hours saved.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Reid Health (regional medical center)
- **Industry:** Hospitals & Health Systems — Regional referral medical center
- **Roles:** Patient Access & Outreach, Revenue Cycle & Prior Auth
- **Tools:** Automated pre-visit registration (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Medicare MSPQ digital completion: 53% [1] (self-reported)
- geriatric engagement: 56% [1] (self-reported)
- staff hours saved: 1600 hours [1] (self-reported)

## How it works
1. Digital MSPQ before the visit, auto-populated from prior answers.
2. Removes downstream EHR work queues.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Reid Health deploys automated pre-visit registration in just four weeks" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/reid-health-deploys-automated-pre-visit-registration-in-just-four-weeks>
   > "Reid Health has saved its registration and customer contact center staff 1600 hours"

Canonical: https://agentcasefile.com/cases/reid-health-notable-pre-visit-registration/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Redfin using an AI agent for home search?

> Redfin built a conversational home-search agent with Sierra. Sierra reports that users of conversational search view nearly twice as many listings and are 47% more likely to request a tour.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Redfin (online real estate brokerage)
- **Industry:** Software, Internet & Media — Real estate marketplace
- **Roles:** Sales Development (SDR)
- **Tools:** Sierra conversational search agent (Sierra)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- tour requests: 47% more likely to request a tour [1] (self-reported)

## How it works
1. Agent searches listings using natural-language criteria.
2. Pushes back on unrealistic criteria and suggests alternatives.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "Redfin and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/redfin>
   > "conversational search users view nearly twice as many listings, and are 47% more likely to request a tour."

Canonical: https://agentcasefile.com/cases/redfin-sierra-conversational-home-search/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Rakuten speed up feature delivery with a coding agent?

> Rakuten uses Anthropic's Claude Code, an agentic coding tool, across engineering; in one test it worked autonomously for seven hours on a complex vLLM implementation and achieved 99.9% numerical accuracy versus the reference method. Anthropic reports Rakuten's average time to market for new features fell from 24 working days to 5 (79%).

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Rakuten (technology company)
- **Industry:** Software, Internet & Media — E-commerce & internet services
- **Roles:** Software Engineering
- **Tools:** Claude Code (Anthropic)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- time to market: 24 working days to 5 days (79% reduction) [1] (self-reported)
- autonomous run accuracy: 99.9% numerical accuracy vs reference method [1] (self-reported)

## How it works
1. Engineer assigns a complex task in natural language.
2. Claude Code works through a large codebase (vLLM, 12.5M lines) with occasional guidance.
3. Output is validated against a reference implementation.

## Risks & compliance
- Code review and testing remain required

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Anthropic, "Rakuten accelerates development with Claude Code" (undated; accessed 2026-10-17) <https://claude.com/customers/rakuten>
   > "Engineering teams saw their average time to market for new features drop from 24 working days to 5 days—a 79% reduction that changed how Rakuten delivers innovation."

Canonical: https://agentcasefile.com/cases/rakuten-claude-code-engineering/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is R1 RCM using AI agents for patient billing calls?

> R1 RCM, which runs revenue cycle operations for health systems, is testing a Sierra AI agent on incoming patient calls. Sierra reports the agent resolving up to 40% of incoming calls in early testing.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** R1 RCM (healthcare revenue cycle management company)
- **Industry:** Hospitals & Health Systems — Revenue cycle management (provider services)
- **Roles:** Revenue Cycle & Prior Auth, Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** pilot · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- call resolution (early testing): up to 40% of incoming calls [1] (self-reported)

## How it works
1. AI agent answers inbound patient calls on behalf of R1's provider clients.
2. Escalates unresolved calls to human representatives.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited
- PHI and patient financial data (HIPAA)

## Notes
- Early-testing figure; deployment status recorded as pilot.

## Sources
1. Sierra, "R1 RCM and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/r1-rcm>
   > "The AI agent is already resolving up to 40% of incoming calls in early testing."

Canonical: https://agentcasefile.com/cases/r1-rcm-sierra-patient-billing-voice-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Qualified generate pipeline with 35+ AI agents?

> Qualified built 35+ specialized AI agents with Relevance AI across its revenue operations. Relevance AI reports a further $7 million in pipeline and $500,000 in closed revenue over six months.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Qualified (marketing software company)
- **Industry:** Software, Internet & Media — SaaS (agentic marketing)
- **Roles:** Sales Ops & Quoting, Sales Development (SDR)
- **Tools:** AI agents (Relevance AI)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- pipeline: $7 million in six months [1] (self-reported)
- closed revenue: $500,000 in six months [1] (self-reported)

## How it works
1. Specialized agents cover distinct revenue-operations tasks.
2. Agents run alongside the human go-to-market team.

## Risks & compliance
- Outbound messaging compliance (spam, consent rules)
- Vendor-published results; not independently audited

## Sources
1. Relevance AI, "How Qualified scaled with 35+ AI agents" (undated; accessed 2026-10-17) <https://relevanceai.com/customers/qualified>
   > "The company partnered with Relevance AI to build 35+ specialized AI agents that generated a further $7 million in pipeline and $500,000 in closed revenue over just six months."

Canonical: https://agentcasefile.com/cases/qualified-relevance-ai-35-agents-revops/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Qonto using AI agents for small-business financial admin?

> Qonto's AI agents draft bank transfers, invoices, and pay slips for small-business customers, who approve before anything is sent. Anthropic reports transfers 2X faster, invoice creation 3X faster, and payroll 5X quicker.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Qonto (business banking fintech)
- **Industry:** Banking & Financial Services — SMB neobank (Europe)
- **Roles:** Customer Support
- **Tools:** Claude on Amazon Bedrock (Anthropic)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- bank transfers: 2X faster [1] (self-reported)
- invoice creation: 3X less time [1] (self-reported)
- payroll runs: 5X quicker [1] (self-reported)

## How it works
1. Agents prepare transfers without manual beneficiary search or reference copying.
2. Agents draft invoices and pay slips.
3. Customer gives final approval before money moves or documents send.

## Risks & compliance
- Vendor-published results; not independently audited
- Payments require explicit customer approval (human in the loop)

## Sources
1. Anthropic, "How Qonto delegates financial admin for small businesses with Claude on Amazon Bedrock" (undated; accessed 2026-10-17) <https://claude.com/customers/qonto>
   > "Cuts client invoice creation time 3X, with its AI agents drafting every invoice and the customer confirming before it's sent"

Canonical: https://agentcasefile.com/cases/qonto-claude-financial-admin-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Swiss bank Pictet using Claude Code?

> Pictet has given around 700 people access to Claude Code and Claude Cowork in a staged rollout and trained more than 500 through 25 workshops. Anthropic reports working prototypes generated in two hours instead of two weeks.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Pictet (private bank / asset manager)
- **Industry:** Banking & Financial Services — Private banking (Switzerland)
- **Roles:** Software Engineering
- **Tools:** Claude Code and Claude Cowork (Anthropic)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- rollout: around 700 people with access; 500+ trained in 25 workshops [1] (self-reported)

## How it works
1. Staged rollout starting with the 1,500-person technology division.
2. Hands-on Claude Code workshops with Artefact.
3. Used to build working prototypes quickly.

## Risks & compliance
- AI-written code or fixes need review before production
- Access to production systems and logs
- Vendor-published results; not independently audited
- Banking secrecy and data-residency constraints

## Sources
1. Anthropic, "Pictet turns weeks of work into hours with Claude Code" (undated; accessed 2026-10-17) <https://claude.com/customers/pictet>
   > "Gave around 700 people access to Claude Code and Claude Cowork to date"

Canonical: https://agentcasefile.com/cases/pictet-claude-code-rollout/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Philippine Airlines using a voice AI agent in its contact center?

> Philippine Airlines deployed a Sierra voice agent in a contact center handling close to two million inbound calls a year. Sierra reports booking retrieval success lifted from 30% to 75% and mis-routed loyalty calls cut from 15.5% to 7.2% after post-launch tuning.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Philippine Airlines (airline)
- **Industry:** Transportation & Logistics — Airline
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- booking retrieval success: from 30% to 75% [1] (self-reported)
- mis-routed loyalty calls: from 15.5% to 7.2% [1] (self-reported)
- FAQ accuracy (pre-launch testing): 98%+ across more than 900 scenarios [1] (self-reported)

## How it works
1. Voice agent handles booking retrieval, FAQs, and loyalty transfers.
2. Post-launch data analysis used to simplify verification and question flows.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "Philippine Airlines and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/philippine-airlines>
   > "The agent delivered 98%+ accuracy across more than 900 FAQ test scenarios and a perfect pass rate on functional UAT."

Canonical: https://agentcasefile.com/cases/philippine-airlines-sierra-voice-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Personio using an AI agent for customer support?

> Personio (16,000 customer companies) uses Fin for support. Intercom reports a ~80% Fin automation rate and a 25-30% reduction in the global human support contact rate in Q1.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Personio (HR software company)
- **Industry:** Software, Internet & Media — HR SaaS (Europe)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin automation rate: ~80% [1] (self-reported)
- human support contact rate: 25–30% reduction in Q1 [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Personio and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/personio>
   > "We had assumed we'd be in the 60 to 70% range."

Canonical: https://agentcasefile.com/cases/personio-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Peddle save with an AI support agent?

> Peddle, an online car-selling marketplace, uses Fin for customer support. Intercom reports $163K in annual cost savings and more than 899 hours saved per month.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Peddle (online car-selling marketplace)
- **Industry:** Consumer Products & Retail — Automotive marketplace
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- annual cost savings: $163K [1] (self-reported)
- hours saved per month: >899 [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "How Peddle saves $163K a year with Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/peddle>
   > "How Peddle saves $163K a year with Fin"

Canonical: https://agentcasefile.com/cases/peddle-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Paychex using AI agents for payroll service calls?

> Paychex uses a Sierra agent for transactional payroll calls such as entering hours. Sierra reports early results of 95% positive client satisfaction and 93% positive customer effort scores.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Paychex (payroll and HR services provider)
- **Industry:** Professional Services — Payroll & HR services
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- client satisfaction: 95% positive [1] (self-reported)
- customer effort score: 93% positive [1] (self-reported)

## How it works
1. Agent handles transactional payroll requests by phone.
2. Service reps focus on complex cases.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited
- Regulated financial disclosures and identity verification

## Sources
1. Sierra, "Paychex and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/paychex>
   > "The early results from those using the agent validated the approach: 95% positive client satisfaction scores, 93% positive customer effort scores"

Canonical: https://agentcasefile.com/cases/paychex-sierra-payroll-voice-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Palo Alto Networks automate demo provisioning with an AI Slack bot?

> A Palo Alto Networks solutions architect built AI Zap workflows in Slack that interpret plain-language requests for demo accounts, provision them across systems, assign licenses, and route approvals. Zapier reports $150K a year saved (1 FTE) and 3,000+ internal users served.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Palo Alto Networks (cybersecurity company)
- **Industry:** Software, Internet & Media — Cybersecurity
- **Roles:** Sales Ops & Quoting
- **Tools:** Zapier (AI Zap workflows) (Zapier)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- savings: $150K per year (1 FTE) [1] (self-reported)
- internal users: 3,000+ [1] (self-reported)

## How it works
1. Employees type requests in plain language in Slack.
2. AI interprets intent, provisions accounts across systems, assigns licenses.
3. Approvals routed to the right people.

## Risks & compliance
- Access provisioning errors; approval controls needed
- Vendor-published results

## MedTech angle
Demo/eval provisioning is analogous to device demo, loaner, and trial-unit logistics.

## Sources
1. Zapier, "How Palo Alto Networks saves $150,000 a year with AI Zap workflows in Slack" (undated; accessed 2026-10-17) <https://zapier.com/customer-stories/palo-alto-networks>
   > "Stas Bilder replaced manual demo provisioning with an AI-powered Slack bot that serves 3,000+ employees"

Canonical: https://agentcasefile.com/cases/palo-alto-networks-zapier-ai-slack-provisioning/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Oviva using Agentforce for patient support?

> Oviva, a digital health provider for weight-related conditions, uses Salesforce Agentforce trained on its FAQs as first-level patient support. Salesforce reports more than 7,000 queries handled in the first two months, about 40% fully resolved by Agentforce.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Oviva (digital health provider (weight-related conditions))
- **Industry:** Hospitals & Health Systems — Digital health / chronic care
- **Roles:** Patient Access & Outreach, Customer Support
- **Tools:** Agentforce (Salesforce); Experience Cloud (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- queries handled (first two months): more than 7,000 [1] (self-reported)
- fully resolved by AI: around 40% [1] (self-reported)

## How it works
1. Agentforce answers patient FAQs on Experience Cloud.
2. Unresolved issues are routed to staff via Omnichannel Routing.

## Risks & compliance
- Health data protection (GDPR); no clinical advice
- Vendor-published results

## Sources
1. Salesforce, "Trailblazer Story: Oviva" (undated; accessed 2026-10-17) <https://www.salesforce.com/eu/blog/trailblazer-oviva/>
   > "In just the first two months, more than 7,000 support queries have been handled this way, with around 40% fully resolved by Agentforce."

Canonical: https://agentcasefile.com/cases/oviva-agentforce-patient-self-service/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Nuuly absorb 40% growth in support with an AI agent?

> Nuuly, URBN's clothing rental subscription, used Fin to handle support through 40% growth in a year without continually hiring. Intercom reports $359k in salary savings in 6 months, with CSAT remaining at 95%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Nuuly (clothing rental subscription service (URBN))
- **Industry:** Consumer Products & Retail — Apparel rental subscription
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- salary savings: $359k in 6 months [1] (self-reported)
- CSAT: 95% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Nuuly and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/nuuly>
   > "The customer experience Natalie set out to protect stayed strong, with CSAT remaining at 95%."

Canonical: https://agentcasefile.com/cases/nuuly-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Numan use an AI agent for patient support?

> Numan, a UK digital health company, uses Fin for patient support. Its team first manually reviewed 100% of Fin interactions, then moved to a 5% monthly audit, while maintaining a 90% CSAT, per Intercom.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Numan (digital health company)
- **Industry:** Hospitals & Health Systems — Digital health (UK)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- CSAT: 90% [1] (self-reported)
- quality assurance: 100% manual review at launch, then 5% monthly audit [1] (self-reported)

## How it works
1. Fin answers patient support questions.
2. Every interaction reviewed manually at first; later a 5% monthly audit.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Health information; clinical questions must go to clinicians

## Sources
1. Intercom (Fin), "Numan and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/numan>
   > "The team began by manually reviewing 100% of Fin interactions to ensure quality, accuracy, and safety."

Canonical: https://agentcasefile.com/cases/numan-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Nubank use Devin to migrate a multi-million-line ETL monolith?

> Nubank delegated the migration of its 8-year-old, multi-million-line core ETL monolith to Devin. Cognition reports a 12x efficiency improvement in engineering hours saved and over 20x cost savings.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Nubank (digital bank)
- **Industry:** Banking & Financial Services — Fintech bank (Latin America)
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- engineering efficiency: 12x improvement in engineering hours saved [1] (self-reported)
- cost savings: over 20x [1] (self-reported)

## How it works
1. Engineers delegate migration sub-tasks to Devin.
2. Engineers review and merge Devin's changes.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "How Nubank refactors millions of lines of code to improve engineering efficiency with Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/nubank>
   > "engineers were able to delegate Devin to handle their migrations and achieve a 12x efficiency improvement in terms of engineering hours saved, and over 20x cost savings."

Canonical: https://agentcasefile.com/cases/nubank-devin-etl-migration/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Novo Nordisk using AI to write clinical study reports?

> Novo Nordisk built NovoScribe, a documentation platform using Claude models on Amazon Bedrock and MongoDB Atlas, to draft clinical study reports, device verification protocols, and patient materials for human review. Anthropic's case study reports documentation time cut from 10+ weeks to 10 minutes, a 95% reduction in resources for device verification protocols, and review cycles cut in half.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Novo Nordisk (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Pharmaceuticals (R&D documentation)
- **Roles:** Regulatory & Medical Writing
- **Tools:** Claude / Claude Code (Anthropic); Amazon Bedrock (AWS); MongoDB Atlas (MongoDB)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- CSR writing time: cut by 90%; clinical study documentation from 10+ weeks to 10 minutes [1] (self-reported)
- device verification protocols: 95% reduction in resources needed [1] (self-reported)

## How it works
1. RAG combines domain-expert-approved text with case-specific variables.
2. Claude drafts the clinical study report (up to 300 pages) or other document.
3. Drafts go to human reviewers for approval.
4. Expanded from CSRs to device verification protocols and patient booklets.

## Risks & compliance
- Regulatory submissions require human sign-off
- Data handling in a GxP context

## MedTech angle
Includes device verification protocols (Novo makes injection pens), relevant to MedTech regulatory and quality teams.

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Anthropic, "Novo Nordisk accelerates clinical documentation and drug development with Claude" (undated; accessed 2026-10-17) <https://claude.com/customers/novo-nordisk>
   > "Claude has helped us cut writing times on CSRs by 90% so we can get documentation directly into human hands for review and approval,"

Canonical: https://agentcasefile.com/cases/novo-nordisk-novoscribe-clinical-documentation/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Notion using AI agents for customer support?

> Notion uses Decagon AI agents across about one million customer inquiries a year. Decagon reports ticket resolution time improved up to 34% and an average ask-for-human rate of 3.4%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Notion (productivity software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- ticket resolution time: improved up to 34% [1] (self-reported)
- ask-for-human rate: 3.4% average [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "Notion and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/notion>
   > "The first major impact of implementing Decagon was ticket resolution time improved up to 34%."

Canonical: https://agentcasefile.com/cases/notion-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did North Kansas City Hospital automate the work of 80 FTEs?

> North Kansas City Hospital (Missouri) automated pre-registration and check-in with Notable. Notable says that by MGMA benchmarks the system would otherwise need 80 additional FTEs, and that patients see a more than 90% reduction in check-in time.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** North Kansas City Hospital (community hospital)
- **Industry:** Hospitals & Health Systems — Community hospital
- **Roles:** Patient Access & Outreach
- **Tools:** Registration and intake automation (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- staffing equivalent: 80 additional FTEs avoided (MGMA benchmark estimate) [1] (self-reported)
- check-in time: more than 90% reduction [1] (self-reported)

## How it works
1. Automated pre-registration for scheduled appointments (previously only 40% could be pre-registered).
2. Digital check-in.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Notable, "NKCH automates 80 FTEs" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/nkch-automates-80-ftes>
   > "Patients experience a more than 90% reduction in check-in time and as a result, have to spend less time waiting."

Canonical: https://agentcasefile.com/cases/north-kansas-city-hospital-notable-80-fte-automation/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Noom using AI agents for member support?

> Noom uses Decagon AI agents for member support. Decagon reports deflection improving from 59% to 68% within 30 days.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Noom (digital health and weight-management company)
- **Industry:** Consumer Products & Retail — Consumer digital health
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- support deflection: from 59% to 68% within 30 days [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "Noom and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/noom>
   > "Within just 30 days, support deflection improved from 59% to 68%, driving greater efficiency and satisfaction."

Canonical: https://agentcasefile.com/cases/noom-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did North Kansas City Hospital double new patient appointments with automated scheduling?

> North Kansas City Hospital and Meritas Health (Kansas City, Missouri) use Notable's automated scheduling. Notable reports scheduled appointments for the same number of providers rising from 5.7% to 14%, described as a 2X increase in new patient appointments.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** North Kansas City Hospital and Meritas Health (health system)
- **Industry:** Hospitals & Health Systems — Community health system
- **Roles:** Patient Access & Outreach
- **Tools:** Intelligent Scheduling (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- new patient appointments (same providers): 5.7% to 14% [1] (self-reported)

## How it works
1. Patients book online; appointments flow into the EHR without staff intervention.
2. Automated registration, check-in, and reminders.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "North Kansas City Hospital and Meritas Health's automated scheduling drives 2X increase in new patient appointments" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/north-kansas-city-hospital-and-meritas-healths-automated-scheduling-drives-2x-increase-in-new-patient-appointments>
   > "Scheduled appointments for the same number of providers jumps from 5.7% to 14%."

Canonical: https://agentcasefile.com/cases/nkch-meritas-notable-automated-scheduling/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does a retina practice use AI voice agents for urgent scheduling?

> Northern California Retina-Vitreous Associates uses an Assort Health voice agent for about 10,000 monthly calls, roughly a third of which were previously missed. The agent applies diagnosis-based urgency rules, such as scheduling retinal detachment within one to two days.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Northern California Retina-Vitreous Associates (retina specialty practice)
- **Industry:** Hospitals & Health Systems — Ophthalmology (retina)
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI agent (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- call volume handled: ~10,000 monthly calls, ~33% previously missed [1] (self-reported)

## How it works
1. Agent answers inbound calls and schedules using diagnosis-based urgency rules.
2. Handles long, complex appointment types and payer requirements.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Assort Health, "NCRVA customer story" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/ncrva>
   > "Assort handles ~10,000 monthly calls with ~33% previously missed, freeing staff to focus on patient care"

Canonical: https://agentcasefile.com/cases/ncrva-retina-assort-voice-ai-scheduling/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is MUSC Health using AI agents for patient access?

> MUSC Health (16 hospitals, South Carolina) uses Notable's AI platform for registration and intake, referrals and scheduling, and care gap closure. Notable reports $3.3M in annual value, 14,500 no-shows avoided, and 5,000+ staff hours reallocated per month.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** MUSC Health (academic health system)
- **Industry:** Hospitals & Health Systems — Academic health system
- **Roles:** Patient Access & Outreach, Revenue Cycle & Prior Auth
- **Tools:** AI platform (intake, referrals, care gaps) (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- annual value: $3.3M [1] (self-reported)
- no-shows avoided: 14,500 [1] (self-reported)
- staff hours reallocated: 5,000+ per month [1] (self-reported)

## How it works
1. Automated digital registration and intake, including touchless copay collection.
2. Referral processing and scheduling automation.
3. Automated care gap outreach.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "How MUSC is using AI to transform access for patients and eliminate manual work for staff" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-story/musc-health>
   > "How MUSC is using AI to transform access for patients and eliminate manual work for staff"

Canonical: https://agentcasefile.com/cases/musc-health-notable-ai-patient-access/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is MPB using an AI agent for customer support?

> MPB, a used camera gear marketplace, uses Fin globally. Intercom reports Fin resolving nearly 10,000 conversations a month, resolution rising from 25-30% to 40-50%, and an 83% Customer Experience Score.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** MPB (used camera gear marketplace)
- **Industry:** Consumer Products & Retail — Re-commerce
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- monthly conversations resolved: nearly 10,000 [1] (self-reported)
- resolution rate: from 25-30% to a consistent 40-50% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "MPB and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/mpb>
   > "Since launch, MPB has doubled its resolution rate, rising from 25-30% to a consistent 40-50%."

Canonical: https://agentcasefile.com/cases/mpb-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# What did Montage Health get from automating intake, referrals, and care gaps?

> Montage Health (California) uses Notable for registration and intake, referrals, and care gap closure. Notable's study reports an estimated $2 million in annual gross value, no-shows down 11 percent, and the equivalent of 13 FTEs freed.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Montage Health (health system)
- **Industry:** Hospitals & Health Systems — Community health system
- **Roles:** Patient Access & Outreach, Revenue Cycle & Prior Auth
- **Tools:** AI platform (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- annual gross value: an estimated $2 million [1] (self-reported)
- no-show rate: dropped by 11 percent [1] (self-reported)
- capacity: equivalent of 13 FTEs freed [1] (self-reported)

## How it works
1. Digital registration and intake with copay collection.
2. Automated referral processing.
3. Automated reminders and care-gap closure outreach.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Montage Health AI Case Study" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-story/montage-health>
   > "The partnership yields Montage Health an estimated $2 million in annual gross value through revenue capture and cost savings."

Canonical: https://agentcasefile.com/cases/montage-health-notable-ai-platform/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Michigan Orthopedic Surgeons use AI agents for patient scheduling?

> Michigan Orthopedic Surgeons deployed EHR-integrated Assort Health AI agents for phone and web scheduling. The vendor reports $2.3 million in additional revenue, +5% appointment volume, and 3.5x higher self-scheduling conversion than its previous provider.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Michigan Orthopedic Surgeons (orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI patient access platform (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- additional revenue: $2.3 million [1] (self-reported)
- appointment volume: +5% total appointment volume growth [1] (self-reported)
- self-scheduling conversion: 3.5x higher vs previous provider [1] (self-reported)

## How it works
1. AI agents integrated with the EHR book, reschedule, cancel, and confirm appointments.
2. Calls needing a human are routed to the right department.
3. Powers web self-scheduling and 24/7 access.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Sources
1. Assort Health, "Michigan Orthopedic Surgeons Generates $2.3M Through AI Patient Scheduling" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/michigan-orthopedic-surgeons-customer-story>
   > "After evaluating 15 potential vendors, they selected Assort Health"

Canonical: https://agentcasefile.com/cases/michigan-orthopedic-surgeons-assort-ai-scheduling/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Medtronic using voice AI agents for medical device support?

> Medtronic runs voice-first AI agents built with Teneo.ai for device support, sales enablement, and internal HR/IT/legal help across 135 countries. Teneo reports $9-10 million saved by the Cardiovascular group, cost per contact cut from $26 to below $12, and 4-5 hours per week saved for sales reps.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Medtronic (medical device manufacturer)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Diversified medical devices
- **Roles:** Customer Support, Field Sales Reps
- **Tools:** Voice-first agentic AI agents (Teneo.ai)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- savings (Cardiovascular group): $9-10 million [1] (self-reported)
- cost per contact: from $26 to below $12 (54% reduction) [1] (self-reported)
- sales rep time: 4-5 hours per week saved, enabling 3 additional sales calls weekly [1] (self-reported)

## How it works
1. 13 active agents: 5 internal (HR, IT, legal) and 8 external (patients, providers, distributors, sales).
2. Device troubleshooting, error-code resolution, multilingual technical support.
3. Sales enablement with instant documentation access.

## Risks & compliance
- Regulated device support; agents must not replace clinical decision-making
- Vendor-published webinar summary; not independently audited

## MedTech angle
Reps reclaiming 4-5 hours a week via instant documentation access is a concrete benchmark for device sales teams.

## Sources
1. Teneo.ai, "Webinar on Demand: How Medtronic Achieves $10M in Savings with Voice-First Agentic AI" (undated; accessed 2026-10-17) <https://www.teneo.ai/blog/webinar-on-demand-compliance-meets-llms-with-medtronic>
   > "Cost per contact reduced from $26 to below $12 – a 54% reduction"

Canonical: https://agentcasefile.com/cases/medtronic-teneo-voice-agents-device-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did MDCS Dermatology use voice AI agents for scheduling?

> MDCS Dermatology (New York and New Jersey) uses Assort Health voice AI agents for patient calls. Assort reports 460 staff hours saved monthly, doubled labor capacity, 29% appointment volume growth supported, and 95% accuracy by the practice's own audit.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** MDCS Dermatology (dermatology practice)
- **Industry:** Hospitals & Health Systems — Dermatology
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI agents (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- staff hours saved: 460 per month [1] (self-reported)
- appointment volume growth supported: 29% [1] (self-reported)
- accuracy (practice audit): 95% [1] (self-reported)

## How it works
1. Agents pre-loaded with specialty scheduling patterns and routing logic.
2. Practice protocols and EHR data converted into live workflows.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Assort Health, "MDCS Dermatology Doubles Labor Capacity and Saves 460 Staff Hours Monthly" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/mdcs--customer-story>
   > "reaching 95% accuracy by the practice's own audit within weeks of go-live"

Canonical: https://agentcasefile.com/cases/mdcs-dermatology-assort-voice-ai/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is L'Oréal using Agentforce in consumer care?

> L'Oréal uses Agentforce for order-status support and in the service console, where agentic AI summarizes cases, drafts emails, and validates intent for advisors. Salesforce reports consumer support cases handled 64% faster; the initial goal is to autonomously resolve 60% of order inquiries.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** L'Oréal (beauty products company)
- **Industry:** Consumer Products & Retail — Beauty & personal care
- **Roles:** Customer Support
- **Tools:** Agentforce (Salesforce)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- case handling speed: 64% faster [1] (self-reported)

## How it works
1. Order status agent answers shoppers around the clock and routes to care teams.
2. In the service console, AI summarizes cases, drafts emails, and generates promo codes.

## Risks & compliance
- Incorrect answers to customers; escalation needed
- Vendor-published results; not independently audited

## Notes
- 60% autonomous resolution is a stated goal, not an outcome.

## Sources
1. Salesforce, "L'Oréal handles consumer support cases 64% faster with Agentforce" (undated; accessed 2026-10-17) <https://www.salesforce.com/customer-stories/loreal/agentic-order-status-support>
   > "L'Oréal handles consumer support cases 64% faster with Agentforce"

Canonical: https://agentcasefile.com/cases/loreal-agentforce-consumer-care/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did LMH Health answer every patient call on the first ring with an AI agent?

> LMH Health (Lawrence, Kansas) deployed a Notable AI agent at its OrthoKansas clinic, where nearly every call had gone to voicemail. Notable reports 100% of 16,000+ inbound calls answered on the first ring, 78.4% of callers engaging with the agent, and 24.3% handled without reaching staff.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** LMH Health (community health system)
- **Industry:** Hospitals & Health Systems — Community health system (orthopedic clinic pilot)
- **Roles:** Patient Access & Outreach
- **Tools:** AI Agent for inbound call triage (Notable)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- first-ring answer rate: 100% of 16,000+ inbound calls [1] (self-reported)
- caller engagement: 78.4% [1] (self-reported)
- handled without staff: 24.3% of inbound calls [1] (self-reported)

## How it works
1. AI agent answers inbound calls and triages them to scheduling, clinical, billing, records, or care teams.
2. Resolves frequently asked questions directly.
3. Foundation for a system-wide AI-enabled call center.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## MedTech angle
OrthoKansas is an orthopedic clinic; a direct analog for ortho device reps' customer practices.

## Sources
1. Notable, "How LMH Health answers every patient call on the first ring" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/how-lmh-health-answers-every-patient-call-on-the-first-ring>
   > "The team was budgeted for 9 FTEs but was chronically short-staffed, and the live answer rate was essentially 0%."

Canonical: https://agentcasefile.com/cases/lmh-health-orthokansas-notable-ai-call-triage/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# What share of support volume does Intercom's Fin resolve at Lightspeed?

> Point-of-sale company Lightspeed rolled out Intercom's Fin AI agent across a multi-region, multilingual contact center. Lightspeed reports Fin resolves 45–65% of support volume across its workspaces and is involved in 99% of conversations.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Lightspeed (commerce software company)
- **Industry:** Consumer Products & Retail — Retail & hospitality POS software
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution rate: 45–65% of support volume [1] (self-reported)
- involvement: Fin involved in 99% of conversations [1] (self-reported)

## How it works
1. Fin answers inbound conversations first across products and languages.
2. Unresolved conversations go to human agents, who use Intercom's Copilot.

## Risks & compliance
- New human agents see fewer basic queries, which affects training (noted by Lightspeed)

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Intercom (Fin), "How Lightspeed achieves up to 65% resolution rate with Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/lightspeed>
   > "Fin is currently resolving between 45-65% of our support volume across our workspaces."

Canonical: https://agentcasefile.com/cases/lightspeed-fin-support-automation/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How much of League's code is written by AI agents?

> League, which builds software for health plans and providers, went company-wide on Claude. Anthropic's Q&A says roughly 98% of League's code is AI-authored and Claude adoption reached 98%, up from 80% at rollout.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** League (healthcare consumer-experience software company)
- **Industry:** Software, Internet & Media — Health tech SaaS (health plans and providers)
- **Roles:** Software Engineering
- **Tools:** Claude Enterprise / Claude Code (Anthropic)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- AI-authored code: roughly 98% [1] (self-reported)
- adoption: 98%, up from 80% [1] (self-reported)

## How it works
1. Company-wide Claude Enterprise rollout.
2. Engineers delegate most code authoring to AI with human review.

## Risks & compliance
- AI-written code or fixes need review before production
- Access to production systems and logs
- Vendor-published results; not independently audited
- Regulated healthcare data in products

## Sources
1. Anthropic, "Case Study: League" (undated; accessed 2026-10-17) <https://claude.com/customers/league-qa>
   > "yet roughly 98% of League's code is AI-authored."

Canonical: https://agentcasefile.com/cases/league-claude-ai-authored-code/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Kalshi automate customer support with an AI agent?

> Kalshi, whose support volume grew from 30,000 tickets in 2024 toward a projected two million by end of 2026, uses Fin. Intercom reports Fin resolves around 80% of conversations end to end.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Kalshi (prediction market exchange)
- **Industry:** Banking & Financial Services — Prediction markets
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- automation rate: around 80% resolved end to end [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Financial-services disclosures and account security

## Sources
1. Intercom (Fin), "How Kalshi automated 80% of customer support with Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/kalshi>
   > "Fin is now involved in nearly every support conversation and resolves around 80% of them end to end, without a human stepping in."

Canonical: https://agentcasefile.com/cases/kalshi-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Itaú using the Devin coding agent across its software lifecycle?

> Itaú Unibanco, Brazil's largest bank, rolled out Devin across its software development lifecycle. Cognition reports a .NET to Java migration of 59 services 6x faster at 5x lower cost, ~70% of scanner-flagged vulnerabilities resolved automatically, and test coverage raised from ~50% to ~90+%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Itaú Unibanco (bank)
- **Industry:** Banking & Financial Services — Bank (Brazil)
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- service migration: 59 services migrated 6x faster at 5x lower cost [1] (self-reported)
- vulnerability remediation: ~70% resolved automatically [1] (self-reported)
- test coverage: from ~50% to ~90+% [1] (self-reported)

## How it works
1. Devin performs code migrations (.NET to Java, SQL Server objects).
2. Automatically fixes vulnerabilities flagged by SonarQube, Fortify, and Veracode.
3. Generates tests to raise coverage.

## Risks & compliance
- Agent-written code requires review
- Vendor-published results

## Sources
1. Cognition (Devin), "How Itaú Deployed AI Across the SDLC at Global-Finance Scale" (undated; accessed 2026-10-17) <https://devin.ai/customers/itau>
   > "~70% of vulnerabilities from SonarQube/Fortify/Veracode resolved automatically."

Canonical: https://agentcasefile.com/cases/itau-devin-sdlc-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Intermountain Healthcare cut patient check-in time with automation?

> Intermountain Healthcare launched Notable's automated digital registration and intake in four weeks. Notable reports a 25% decrease in check-in time and a 96% patient satisfaction rating.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Intermountain Health (health system)
- **Industry:** Hospitals & Health Systems — Integrated health system
- **Roles:** Patient Access & Outreach
- **Tools:** Digital assistants for registration and intake (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- check-in time: 25% decrease [1] (self-reported)
- patient satisfaction: 96% [1] (self-reported)

## How it works
1. Digital assistants perform repetitive registration tasks the way staff would.
2. Patients complete forms without an app or login; prior data pre-populates.
3. Insurance cards and consents captured into the EHR automatically.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Intermountain Healthcare reduces patient check-in time by 25%" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/intermountain-healthcare-reduces-patient-check-in-time-by-25>
   > "How the leading health system launched a fully automated digital patient experience in just four weeks, which led to a 96% patient satisfaction rating."

Canonical: https://agentcasefile.com/cases/intermountain-notable-digital-check-in/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Impel using Claude agents to ramp sales reps and prep business reviews?

> Impel, which builds AI for auto dealers, uses Claude-based agent workflows for rep training and account work. Anthropic reports new-product ramp time cut from 3 to 6 months to about 1 month, monthly business review prep cut from 3 to 4 hours to about 30 minutes, and 20% more support cases handled.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Impel (automotive AI software company)
- **Industry:** Software, Internet & Media — Automotive SaaS
- **Roles:** Sales Ops & Quoting, Customer Support, Field Sales Reps
- **Tools:** Claude (incl. Claude Code) (Anthropic)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- rep ramp time: from 3 to 6 months to an average of 1 month [1] (self-reported)
- business review prep: from 3 to 4 hours to about 30 minutes [1] (self-reported)
- support capacity: 20% more support cases [1] (self-reported)

## How it works
1. Roleplay and certification agent trains customer-facing reps on new products.
2. Claude drafts monthly business review briefs from Snowflake data.
3. Reclaimed support time goes to migrations and projects.

## Risks & compliance
- Vendor-published results; not independently audited

## MedTech angle
Rep ramp on new products is a constant MedTech problem; AI roleplay certification is a transferable pattern.

## Sources
1. Anthropic, "Impel case study" (undated; accessed 2026-10-17) <https://claude.com/customers/impel>
   > "Reduced monthly business review prep from 3 to 4 hours to about 30 minutes with Claude-drafted briefs from Snowflake"

Canonical: https://agentcasefile.com/cases/impel-claude-gtm-enablement-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is IG Group using an AI agent for trading customer support?

> IG Group uses Fin for customer support. Intercom quotes IG reporting chat deflection of up to 70%, an 80% Fin resolution rate, and "millions of dollars" in cost savings.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** IG Group (online trading platform)
- **Industry:** Banking & Financial Services — Online trading / brokerage
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chat deflection: up to 70% [1] (self-reported)
- Fin resolution rate: 80% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Financial-services disclosures and account security

## Sources
1. Intercom (Fin), "IG Group and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/ig-group>
   > "Some of our deflection rates on chat have reached highs of up to 70%, which is really great for us."

Canonical: https://agentcasefile.com/cases/ig-group-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How are Hunter Douglas's AI agents driving revenue?

> Hunter Douglas uses Decagon AI agents across markets. Decagon reports 40% average deflection, more than $1 million in revenue from conversations handled fully by AI, and 85% higher average order value among customers who engaged an agent.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Hunter Douglas (window coverings manufacturer)
- **Industry:** Consumer Products & Retail — Home furnishings manufacturer
- **Roles:** Customer Support, Sales Development (SDR)
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- revenue from AI-only conversations: more than $1 million [1] (self-reported)
- average order value (engaged vs not): 85% higher [1] (self-reported)
- deflection: 40% of inbound volume on average [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "Hunter Douglas and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/hunter-douglas>
   > "Additionally, more than $1 million in revenue has already come from conversations fully handled by AI, without ever being escalated to a human."

Canonical: https://agentcasefile.com/cases/hunter-douglas-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is HubSpot using Claude agents in engineering and support?

> HubSpot uses Claude, including Claude Code, across engineering and customer success. Anthropic reports up to 40% productivity increases in web development and content workflows and complex troubleshooting cut from 3-5 days to under an hour.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** HubSpot (CRM software company)
- **Industry:** Software, Internet & Media — SaaS (CRM)
- **Roles:** Software Engineering, Customer Support
- **Tools:** Claude Code / Claude API (Anthropic)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- productivity: up to 40% increase [1] (self-reported)
- complex troubleshooting: from 3-5 days to under an hour [1] (self-reported)

## How it works
1. Engineers use Claude Code for migrations and maintenance.
2. Customer success managers use Claude to troubleshoot escalated workflows directly.

## Risks & compliance
- AI-written code or fixes need review before production
- Access to production systems and logs
- Vendor-published results; not independently audited

## Sources
1. Anthropic, "HubSpot Claude Platform (API) case study" (undated; accessed 2026-10-17) <https://claude.com/customers/hubspot>
   > "Reduced complex technical troubleshooting from 3-5 days to under an hour"

Canonical: https://agentcasefile.com/cases/hubspot-claude-code-engineering-troubleshooting/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Homeworks raise chat resolution with an AI agent?

> Homeworks moved from keyword-rule bots that resolved only 20 to 30% of chats to Intercom's Fin. Intercom reports an 85% Fin resolution rate and 92% involvement.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Homeworks (home services company)
- **Industry:** Consumer Products & Retail — Home services
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 85% [1] (self-reported)
- prior bot resolution: 20 to 30% of chats [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Homeworks and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/homeworks>
   > "every reply had to be built manually with keyword rules, and the bot resolved only 20 to 30% of chats."

Canonical: https://agentcasefile.com/cases/homeworks-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Henry Schein Australia using Agentforce for customer chats?

> Henry Schein Australia, a dental and medical supply distributor, deployed Salesforce Agentforce for customer chats, most of which were order questions. Its implementation partner reports Agentforce now handles 70% of incoming enquiries.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Henry Schein Australia (dental and medical supply distributor)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Dental & medical supply distribution
- **Roles:** Customer Support
- **Tools:** Agentforce Service Agent (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- enquiries handled by AI: 70% [1] (self-reported)

## How it works
1. Agentforce answers "Where's my order?" and order-change chats.
2. Complex interactions handed to human agents.

## Risks & compliance
- Order and pricing accuracy for regulated healthcare products
- Partner-published results; not independently audited

## MedTech angle
Order-status agents for distributors mirror what device companies field from hospitals and ASCs.

## Sources
1. ProQuest IT (implementation partner), "70% of Customer Chats Now Handled by AI – Agentforce Transforms Henry Schein’s Service" (undated; accessed 2026-10-17) <https://proquestit.com/insights/henry-schein-agentforce-service-chats/>
   > "Today, Agentforce handles 70% of incoming enquiries"

Canonical: https://agentcasefile.com/cases/henry-schein-australia-agentforce-order-chats/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Good Shepherd Rehabilitation reduce no-shows with automation?

> Good Shepherd Rehabilitation Network replaced paper reminder cards and staff reminder calls with Notable automated reminders and digital intake. Notable reports no-shows down 32% and a 92% patient satisfaction rating.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Good Shepherd Rehabilitation Network (rehabilitation provider)
- **Industry:** Hospitals & Health Systems — Physical rehabilitation
- **Roles:** Patient Access & Outreach
- **Tools:** Digital patient experience automation (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- no-shows: reduced by 32% [1] (self-reported)
- patient satisfaction: 92% [1] (self-reported)

## How it works
1. Automated reminders for multi-visit therapy schedules.
2. Digital intake before appointments.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## MedTech angle
Post-surgical rehab adherence matters to orthopedic outcomes.

## Sources
1. Notable, "Good Shepherd Rehabilitation Network delivers best-in-class digital patient experience; reduces no-shows by 32%" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/good-shepherd-rehabilitation-network-delivers-best-in-class-digital-patient-experience-reduces-no-shows-by-32>
   > "Notable had an immediate impact on this front; the 92% patient satisfaction rating speaks for itself."

Canonical: https://agentcasefile.com/cases/good-shepherd-rehab-notable-no-show-reduction/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is GoFundMe using an AI agent to coach fundraisers?

> GoFundMe built a Sierra AI agent that coaches fundraising organizers. Sierra reports organizers raised 5% more in donations in early results, which could mean an estimated $125 million in additional help raised in the U.S.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** GoFundMe (online fundraising platform)
- **Industry:** Software, Internet & Media — Crowdfunding platform
- **Roles:** Customer Support
- **Tools:** Sierra AI coaching agent (Sierra)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- donations raised: 5% more; estimated $125 million additional (projection) [1] (self-reported)

## How it works
1. Agent offers contextual guidance drawn from millions of past fundraisers.
2. Tested with simulations and red-team exercises before launch.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Notes
- $125 million is the vendor's projection from early results, not a realized figure.

## Sources
1. Sierra, "GoFundMe and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/gofundme>
   > "Organizers raised 5% more in donations, which - based on early results - could result in an estimated $125 million in additional help raised in the U.S."

Canonical: https://agentcasefile.com/cases/gofundme-sierra-fundraiser-coaching-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Glean using an AI agent for customer support?

> Glean uses Fin across its help center, developer docs, community, and support portal. Intercom reports Fin handling 100% of chats within three months, resolving 83% without a human, with 95% customer satisfaction.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Glean (enterprise AI search company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chat automation rate: 83% [1] (self-reported)
- customer satisfaction: 95% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Glean and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/glean>
   > "83% of these chats were resolved by Fin without a human needing to step in, with a high customer satisfaction rate of 95%."

Canonical: https://agentcasefile.com/cases/glean-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Gillette Children's automate scheduling and intake?

> Gillette Children's uses Notable so families can self-schedule at any hour and complete intake digitally. Notable reports 118 staff hours saved per week, about 99% satisfaction with intake and 95% with scheduling, and 700 RN minutes a month saved by one pre-procedure form.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Gillette Children's (pediatric specialty hospital)
- **Industry:** Hospitals & Health Systems — Pediatric specialty care
- **Roles:** Patient Access & Outreach
- **Tools:** Automated scheduling and intake (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- staff time saved: 118 hours per week [1] (self-reported)
- satisfaction: around 99% intake, 95% scheduling [1] (self-reported)
- RN time: 700 minutes a month from one automated pre-procedure form [1] (self-reported)

## How it works
1. Families book appointments digitally with no staff intervention.
2. Automated intake and pre-procedure forms.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Notable, "Gillette Children's" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/gillette-childrens>
   > "Ultimately, automating intake has saved Gillette 118 hours per week."

Canonical: https://agentcasefile.com/cases/gillette-childrens-notable-scheduling-intake/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Gamma using an AI agent for support at 50 million users?

> Gamma, with a user base past 50 million, uses Fin on all inbound support. Intercom reports Fin involved in 100% of conversations and resolving 75% end-to-end.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Gamma (AI presentation software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 75% [1] (self-reported)
- Fin involvement rate: 100% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Gamma and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/gamma>
   > "Since its implementation, Fin is now involved in 100% of inbound support conversations and resolves 75% of them end-to-end."

Canonical: https://agentcasefile.com/cases/gamma-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Fundrise use an AI agent for investor support?

> Fundrise uses Fin for investor support. Intercom reports Fin handling 50% of total support cases with response accuracy above 95%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Fundrise (real estate investment platform)
- **Industry:** Banking & Financial Services — Online investing
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- support cases handled: 50% [1] (self-reported)
- response accuracy: > 95% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Financial-services disclosures and account security

## Sources
1. Intercom (Fin), "Investing in success: How Fundrise resolves 50% of its support volume with Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/fundrise>
   > "Investing in success: How Fundrise resolves 50% of its support volume with Fin"

Canonical: https://agentcasefile.com/cases/fundrise-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Fort HealthCare automate prior authorizations for its surgery center?

> Fort HealthCare (Wisconsin) automated prior authorizations for its ambulatory surgery center and digitized registration with Notable. Notable reports a 91% prior authorization success rate, a 74% digital registration completion rate, and a more than 5% reduction in cancellations.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Fort HealthCare (community health system)
- **Industry:** Hospitals & Health Systems — Community health system (ambulatory surgery center)
- **Roles:** Revenue Cycle & Prior Auth, Patient Access & Outreach
- **Tools:** Prior authorization and intake automation (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- prior authorization success rate: 91% [1] (self-reported)
- digital registration completion: 74% [1] (self-reported)
- cancellations: more than 5% reduction [1] (self-reported)

## How it works
1. Automation submits and tracks prior authorizations for surgery center cases.
2. Digital registration completed by patients before visits.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## MedTech angle
Ambulatory surgery center prior auth directly gates elective orthopedic and device procedures.

## Sources
1. Notable, "Fort HealthCare automates prior authorizations with 91% success rate" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/fort-healthcare-automates-prior-authorizations-with-91-success-rate>
   > "By digitizing and automating the registration process with Notable, Fort HealthCare has seen more patients successfully complete their registration, reporting a 74% digital completion rate."

Canonical: https://agentcasefile.com/cases/fort-healthcare-notable-prior-authorization/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did FIFA use an AI agent to support media at the World Cup 2026?

> FIFA deployed an Agentforce FAQ agent for journalists, broadcasters, and photographers at the FIFA World Cup 2026. Salesforce reports 3,600+ chat sessions and an 85% deflection rate in the first 45 days, recovering an estimated 1,000 service rep hours.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** FIFA (international sports federation)
- **Industry:** Software, Internet & Media — Sports & media
- **Roles:** Customer Support
- **Tools:** Agentforce (Salesforce)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chat sessions (45 days): more than 3,600 [1] (self-reported)
- deflection rate: 85% [1] (self-reported)
- service rep hours recovered: estimated 1,000 [1] (self-reported)

## How it works
1. Agent answers common media questions instantly.
2. Went live two months before the opening match.

## Risks & compliance
- Incorrect answers to customers; escalation needed
- Vendor-published results; not independently audited

## Sources
1. Salesforce, "FIFA media support with Agentforce" (undated; accessed 2026-10-17) <https://www.salesforce.com/customer-stories/fifa/agentic-media-support>
   > "By answering common questions instantly, in that same 45-day period, the agent achieved an 85% deflection rate and recovered an estimated 1,000 service rep hours."

Canonical: https://agentcasefile.com/cases/fifa-agentforce-media-faq-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did an independent primary care practice use an AI agent to fix Monday call surges?

> Family Practice Associates of Lexington (21 providers, Kentucky) uses an Assort Health AI agent named "Sophie" for phone scheduling, including after hours. The vendor reports a 30%+ increase in new patient visits and zero missed emergent-triage transfers.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Family Practice Associates of Lexington (independent primary care practice)
- **Industry:** Hospitals & Health Systems — Primary care (independent practice)
- **Roles:** Patient Access & Outreach
- **Tools:** AI patient access agent ("Sophie") (Assort Health); Veradigm EHR integration (Veradigm)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- new patient visits: 30%+ increase [1] (self-reported)
- emergent-triage transfers missed: Zero [1] (self-reported)

## How it works
1. AI agent answers calls and books appointments 24/7, integrated with the Veradigm EHR.
2. Encodes the practice's scheduling rules previously held by veteran staff.
3. Transfers emergent-triage calls to humans.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## Sources
1. Assort Health, "Family Practice Associates of Lexington Ends the Monday Morning Phone Rush" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/family-practice-associates-of-lexington-customer-story>
   > "FPA averages 225,000 incoming calls a year."

Canonical: https://agentcasefile.com/cases/family-practice-associates-lexington-assort-ai-scheduling/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is AstraZeneca's Evinova using an AI agent to draft regulatory software documentation?

> Evinova, a health tech company in the AstraZeneca group, uses Devin to draft regulatory documents for its clinical-trial software modules, work that took 35 to 40 hours of cross-functional time. Cognition reports structured first drafts at roughly 90% accuracy, with senior engineers shifting to reviewers.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Evinova (AstraZeneca group) (health tech company (clinical development software))
- **Industry:** Pharma, Biotech & Life Sciences — Clinical development software
- **Roles:** Regulatory & Medical Writing, Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- first-draft accuracy: roughly 90% [1] (self-reported)
- prior effort per module: 35 to 40 hours of cross-functional coordination [1] (self-reported)

## How it works
1. Devin assembles context from code and systems across teams.
2. Produces structured first drafts of regulatory documents.
3. Senior engineers review and approve.

## Risks & compliance
- Regulated (GxP) software documentation requires qualified human review
- Vendor-published results

## Sources
1. Cognition (Devin), "Evinova and Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/evinova>
   > "Devin produced structured first drafts at roughly 90% accuracy, shifting senior engineers from primary authors to reviewers"

Canonical: https://agentcasefile.com/cases/evinova-astrazeneca-devin-regulatory-docs/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does EvenUp use AI to draft personal-injury case documents?

> EvenUp's Claude-powered platform drafts personal-injury case documents for law firms. Anthropic reports drafting time cut from 8 to 15 hours of skilled work to minutes, and settlement offers raised 300% for firms presenting the full damages picture.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** EvenUp (legal AI company (personal injury))
- **Industry:** Professional Services — Legal tech (personal injury)
- **Roles:** Regulatory & Medical Writing
- **Tools:** Claude models (Anthropic)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- drafting time: from 8 to 15 hours to minutes [1] (self-reported)
- settlement offers: 300% increase [1] (self-reported)

## How it works
1. Models review case records and draft demand documents with cited evidence.
2. Attorneys review and finalize.

## Risks & compliance
- Vendor-published results; not independently audited
- Legal accuracy; attorney review required
- Handles medical records (PHI)

## Sources
1. Anthropic, "EvenUp cuts document drafting from 15 hours to 15 minutes with Claude" (undated; accessed 2026-10-17) <https://claude.com/customers/evenup>
   > "Raised settlement offers 300% for firms presenting the full damages picture with cited evidence"

Canonical: https://agentcasefile.com/cases/evenup-claude-demand-letter-drafting/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is DXC using AI agents in insurance claims processing?

> DXC Technology, which runs core insurance systems for 1,100 customers, brought Claude into its Assure orchestration platform. Anthropic reports claims document backlogs cut from days to minutes, documents classified in under a second, and new regulatory rules embedded in days instead of 12 to 18 months.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** DXC Technology (IT services company (insurance platforms))
- **Industry:** Banking & Financial Services — Insurance technology
- **Roles:** Customer Support
- **Tools:** Claude (Anthropic); Assure orchestration platform (DXC Technology)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- regulatory rule implementation: days vs 12 to 18 months historically [1] (self-reported)
- first-pass accuracy: 80% right on the first pass [1] (self-reported)

## How it works
1. Claude classifies incoming claims documents.
2. Helps implement new regulatory rules and benefit calculations.

## Risks & compliance
- Vendor-published results; not independently audited
- Regulated insurance calculations require validation

## Sources
1. Anthropic, "DXC brings Claude to the insurance backbone running billions of policies" (undated; accessed 2026-10-17) <https://claude.com/customers/dxc>
   > "Cut claims document backlogs from days to minutes, with documents classified in under a second"

Canonical: https://agentcasefile.com/cases/dxc-claude-insurance-claims-documents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is DaySmart using an AI agent for after-hours support?

> DaySmart added Fin to cover support outside fixed hours. Intercom reports Fin resolves 72% of conversations it touches, handles more than 10,000 resolutions a month, and automates 38% of total support volume.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** DaySmart (business management software company)
- **Industry:** Software, Internet & Media — SMB SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: 72% [1] (self-reported)
- monthly resolutions: more than 10,000 [1] (self-reported)
- share of total support volume automated: 38% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "DaySmart and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/daysmart>
   > "Today, Fin resolves 72% of the conversations it touches, handles more than 10,000 resolutions a month, and automates 38% of total support volume."

Canonical: https://agentcasefile.com/cases/daysmart-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Databox use an AI agent to support revenue growth?

> Databox uses Fin for customer support. Intercom reports a 55% Fin deflection rate, a 50% increase in team productivity, and 40% more revenue, which Intercom attributes to Fin.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Databox (business analytics software company)
- **Industry:** Software, Internet & Media — SaaS (analytics)
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin deflection rate: 55% [1] (self-reported)
- team productivity: 50% increase [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Revenue attribution to the agent is the vendor's framing

## Sources
1. Intercom (Fin), "Databox and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/databox>
   > "Databox generates 40% more revenue thanks to Fin"

Canonical: https://agentcasefile.com/cases/databox-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Curology using AI agents for patient support?

> Curology, a D2C prescription skincare company that has treated nearly 6 million patients, uses Decagon AI agents for support. Decagon reports chat rising from 5% to 80% of ticket volume and support operation costs cut by 65%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Curology (direct-to-consumer telehealth dermatology company)
- **Industry:** Hospitals & Health Systems — Telehealth dermatology (D2C)
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- chat share of tickets: from 5% to 80% [1] (self-reported)
- support operation costs: reduced by 65% [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Health information; clinical questions must route to clinicians

## Sources
1. Decagon, "Curology and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/curology>
   > "Since implementing Decagon, Curology has significantly reduced customer support operation costs by 65%."

Canonical: https://agentcasefile.com/cases/curology-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Contractor Appointments use an AI scheduler to book appointments from texts?

> Contractor Appointments built an AI scheduler with Zapier that parses homeowners' after-hours text replies, checks availability, and auto-books appointments with no human needed. Zapier reports 20 to 50 extra appointments booked automatically per day and $300k in incremental annual revenue.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Contractor Appointments (home-improvement appointment-setting agency)
- **Industry:** Professional Services — Lead generation & appointment setting
- **Roles:** Sales Development (SDR)
- **Tools:** Zapier (AI Zap workflows) (Zapier)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- extra appointments: 20 to 50 per day [1] (self-reported)
- incremental annual revenue: $300k [1] (self-reported)

## How it works
1. AI parses inbound SMS replies to estimates.
2. Checks calendar availability.
3. Books appointments automatically.

## Risks & compliance
- SMS consent (TCPA) and booking errors
- Vendor-published results

## Sources
1. Zapier, "How Contractor Appointments books $134M in client revenue with AI-powered automation" (undated; accessed 2026-10-17) <https://zapier.com/customer-stories/contractor-appointments>
   > "20 to 50 extra appointments are automatically booked daily, capturing late-night replies that previously slipped through the cracks."

Canonical: https://agentcasefile.com/cases/contractor-appointments-zapier-ai-scheduler/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Consensys (MetaMask) using an AI agent for support?

> Consensys, maker of MetaMask, uses Fin for support. Intercom reports Fin resolving nearly 70% of support conversations within 8 weeks, with 90% involvement.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Consensys (Web3 software company (MetaMask))
- **Industry:** Software, Internet & Media — Web3 / crypto wallets
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution within 8 weeks: nearly 70% [1] (self-reported)
- Fin involvement rate: 90% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- Financial-services disclosures and account security

## Sources
1. Intercom (Fin), "Consensys and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/consensys>
   > "Within 8 weeks, Fin was resolving nearly 70% of support conversations, with the CSAT to match."

Canonical: https://agentcasefile.com/cases/consensys-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Concord Orthopaedics automate referrals and cut hold times with AI agents?

> Concord Orthopaedics (New Hampshire; 50+ providers) uses Assort Health inbound and outbound AI agents for calls and referrals. The vendor reports 84% of referrals automated end-to-end, a 68% referral-to-appointment conversion rate, over 95% scheduling accuracy, and hold times near zero.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Concord Orthopaedics (independent orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics
- **Roles:** Patient Access & Outreach
- **Tools:** Inbound and outbound AI agents ("Cheryl") (Assort Health)
- **Status:** production · **Agent pattern:** multi-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- referral automation: 84% of referrals automated end-to-end [1] (self-reported)
- referral conversion: 68% referral-to-appointment [1] (self-reported)
- scheduling accuracy: over 95% of the 720 appointments per week [1] (self-reported)

## How it works
1. Inbound agent ("Cheryl") answers patient calls and schedules.
2. Referral triage matches each referral to the right provider.
3. Outbound agent contacts referred patients and schedules them on the same call.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Notes
- Appointments-per-week figure is asterisked in the vendor story.

## Sources
1. Assort Health, "Concord Orthopaedics Cuts Hold Times and Automates 84% of Referrals" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/concord-orthopaedics-customer-story>
   > "Over 95% of the 720 appointments per week* are scheduled accurately."

Canonical: https://agentcasefile.com/cases/concord-orthopaedics-assort-referral-automation/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did CommonSpirit Health automate Medicare Annual Wellness Visit outreach?

> CommonSpirit Health ran automated multi-touch Medicare Annual Wellness Visit outreach and scheduling with Notable across California, Arizona, and Nevada markets. Notable reports 160,000+ outreaches in 2024, 74,878 staff hours saved (~36 FTE), and 445,888 manual calls avoided.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** CommonSpirit Health (health system)
- **Industry:** Hospitals & Health Systems — Multi-state health system
- **Roles:** Patient Access & Outreach
- **Tools:** Outreach automation and automated scheduling (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- outreach volume: 160,000+ MAWV outreaches in 2024 [1] (self-reported)
- staff hours saved: 74,878 (~36 FTE) [1] (self-reported)
- manual calls avoided: 445,888 [1] (self-reported)

## How it works
1. Market-specific multi-touch outreach cadence (typically six contacts per patient).
2. Patients self-schedule; follow-up centralized in a Patient Communication Center.
3. MAWV prioritized as 50% of care gap outreach.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "CommonSpirit Health: Automated MAWV Outreach at Scale" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/commonspirit-health>
   > "74,878 staff hours saved and CY24 Annual Wellness Visit targets met in under nine months with automated outreach and scheduling"

Canonical: https://agentcasefile.com/cases/commonspirit-notable-mawv-outreach/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Clay using an AI agent for support?

> Clay uses Fin across multiple support channels. Intercom reports Fin involved in 90% of conversations and resolving up to 50%.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Clay (go-to-market data software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin involvement rate: 90% [1] (self-reported)
- Fin resolution rate: up to 50% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "Clay and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/clay>
   > "Now, Fin is enabling Clay to support customers on multiple channels, is involved in 90% of support conversations – and resolving up to 50% of them."

Canonical: https://agentcasefile.com/cases/clay-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did ClassPass scale chat support to 24/7 with AI agents?

> ClassPass deployed Decagon AI agents in chat. ClassPass reports 10x higher deflection at launch than anticipated, letting it expand chat from 16 hours 5 days a week to 24/7.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** ClassPass (fitness and wellness booking marketplace)
- **Industry:** Consumer Products & Retail — Consumer subscription marketplace
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- deflection vs plan: 10x higher at launch than anticipated [1] (self-reported)
- chat hours: from 16 hours 5 days a week to 24 hours 7 days a week [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "ClassPass and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/classpass>
   > "we saw 10x higher deflection at launch than we anticipated."

Canonical: https://agentcasefile.com/cases/classpass-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Circle Medical using an AI agent for patient support?

> Circle Medical, a virtual primary care provider, uses Intercom's Fin AI agent for patient support. Intercom reports a 58% reduction in resolution time and Fin involvement in 96% of conversations.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Circle Medical (telehealth primary care provider)
- **Industry:** Hospitals & Health Systems — Virtual primary care
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution time: 58% reduction [1] (self-reported)
- Fin involvement rate: 96% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited
- PHI handling (HIPAA); must not give clinical advice

## Sources
1. Intercom (Fin), "Circle Medical and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/circle-medical>
   > "We also reduced resolution time by 58%."

Canonical: https://agentcasefile.com/cases/circle-medical-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Chime using voice AI agents at a million calls a month?

> Chime uses Decagon voice AI agents for member calls. Decagon reports scaling to more than 1 million calls per month and consistently seeing 70%+ resolution.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Chime (fintech / consumer banking company)
- **Industry:** Banking & Financial Services — Consumer fintech
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- call volume: more than 1 million calls per month [1] (self-reported)
- resolution: 70%+ [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited
- Account security and identity verification

## Sources
1. Decagon, "Chime and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/chime>
   > "Scaled to more than 1 million calls per month with no reliability issues"

Canonical: https://agentcasefile.com/cases/chime-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did an FQHC use AI voice agents to expand patient access?

> Chesapeake Health Care, a Maryland FQHC with 150+ providers, uses Assort Health voice agents for inbound calls after 25% of calls were dropped. Assort reports $1M+ generated.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Chesapeake Health Care (federally qualified health center)
- **Industry:** Hospitals & Health Systems — FQHC
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI patient access (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- revenue generated: $1M+ [1] (self-reported)

## How it works
1. Voice agents answer inbound calls and schedule appointments.
2. Routes complex calls to staff.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Assort Health, "Chesapeake Health Care Generates $1M+ and Expands Patient Access" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/chesapeake-health-care-customer-story>
   > "For patients, that meant they often waited on hold – sometimes more than 7 minutes – only to be sent to voicemail."

Canonical: https://agentcasefile.com/cases/chesapeake-health-care-assort-patient-access/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Catholic Health use voice AI for patient portal support calls?

> Catholic Health deployed Notable voice AI for MyChart support calls that had relied on outsourced live agents. Notable reports call containment rising from 30% to 57%, a 50% cut in outsourced call volume, and a projected $350,000 a year in savings.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Catholic Health (health system)
- **Industry:** Hospitals & Health Systems — Health system (Buffalo, NY)
- **Roles:** Customer Support, Patient Access & Outreach
- **Tools:** Voice AI for MyChart support (Notable)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- call containment: from 30% to 57% [1] (self-reported)
- outsourced call volume: 50% reduction [1] (self-reported)
- cost savings: on track to save $350,000 a year [1] (self-reported)

## How it works
1. Voice AI answers MyChart (patient portal) support calls.
2. Resolves routine issues; transfers the rest.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Notable, "Catholic Health: AI MyChart Support, 57% Containment" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/catholic-health>
   > "Call containment rates improved dramatically, rising from 30% to 57%, resulting in a 50% reduction in outsourced call center volume."

Canonical: https://agentcasefile.com/cases/catholic-health-notable-voice-ai-mychart-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Castell cut chart review time with AI automation?

> Castell, Intermountain's value-based care company, uses Notable automation to review charts for care-gap closure opportunities. Notable reports average daily chart time falling from four hours to one hour and 1.68 minutes saved per chart.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Castell (Intermountain Health) (value-based care services company)
- **Industry:** Hospitals & Health Systems — Value-based care
- **Roles:** Clinicians & Clinical Ops
- **Tools:** Intelligent Population Health (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- daily chart time: from four hours to one hour [1] (self-reported)
- time saved per chart: 1.68 minutes [1] (self-reported)

## How it works
1. Automation pre-reviews charts and flags care gap closure opportunities.
2. Staff review the flagged items.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Castell reduces average daily chart time from four hours to one hour" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/castell-reduces-average-daily-chart-time-from-four-hours-to-one-hour>
   > "How the value-based care health services company is using intelligent automation to review charts and identify opportunities for care gap closure."

Canonical: https://agentcasefile.com/cases/castell-notable-chart-review-care-gaps/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Carvana using an AI agent to investigate production alerts?

> Carvana uses Claude Tag in Slack to start investigating production alerts, identify the owning team, and post root-cause analyses before an engineer picks them up. Anthropic reports alerts in one channel down 56% and answers 65% faster for its wholesale platform team.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Carvana (online used-car retailer)
- **Industry:** Consumer Products & Retail — E-commerce (automotive)
- **Roles:** Software Engineering
- **Tools:** Claude Tag (Anthropic)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- alert volume (retail science channel): declined 56% [1] (self-reported)
- time to answers (wholesale platform team): 65% faster [1] (self-reported)

## How it works
1. Agent starts investigating when an alert reaches Slack.
2. Identifies the owning team and posts a root-cause analysis.
3. Checks every release against a log baseline from the hour before it shipped.

## Risks & compliance
- AI-written code or fixes need review before production
- Access to production systems and logs
- Vendor-published results; not independently audited

## Sources
1. Anthropic, "Carvana turns Slack alerts into production fixes with Claude Tag" (undated; accessed 2026-10-17) <https://claude.com/customers/carvana>
   > "In the retail science channel, alert volume declined 56% as Claude Tag helped teams address recurring root causes rather than repeatedly clearing the same alerts."

Canonical: https://agentcasefile.com/cases/carvana-claude-tag-alert-root-cause/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Care New England cut authorization-related write-offs with automation?

> Care New England automated radiology prior authorizations and notices of admission with Notable. Notable reports a 55% reduction in authorization-related write-offs and 2,841 staff hours saved; CNE had estimated it would otherwise need 14 more FTEs.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Care New England (health system)
- **Industry:** Hospitals & Health Systems — Regional health system
- **Roles:** Revenue Cycle & Prior Auth
- **Tools:** Prior authorization and NOA automation (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- authorization-related write-offs: 55% reduction [1] (self-reported)
- staff hours saved: 2,841 hours [1] (self-reported)
- hiring avoided (pre-project estimate): 14 FTEs [1] (self-reported)

## How it works
1. Automation completes radiology prior authorizations and notices of admission.
2. Staff handle exceptions.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Notable, "Care New England reports a 55% reduction in authorization-related write-offs" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/care-new-england-reports-a-55-reduction-in-authorization-related-write-offs-since-automating-radiology-prior-authorizations-and-notice-of-admissions>
   > "As a result, CNE reports a 55% reduction in authorization-related write-offs, 2,841 hours saved for staff, and more."

Canonical: https://agentcasefile.com/cases/care-new-england-notable-radiology-prior-auth/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Boston Bone & Joint Institute use AI agents for inbound and outbound patient access?

> Boston Bone & Joint Institute (41 providers) uses Assort Health AI agents for inbound calls and automated outbound rescheduling. The vendor reports call abandonment falling from 9% to 3%, a 53% rescheduling rate from outbound campaigns, and attrition of 2 staff absorbed without backfill.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Boston Bone & Joint Institute (orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics
- **Roles:** Patient Access & Outreach
- **Tools:** Assort Concierge AI agents (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- call abandonment: from 9% to 3% [1] (self-reported)
- automated rescheduling: 53% rescheduling rate from outbound campaigns [1] (self-reported)
- staffing: attrition of 2 staff members absorbed without backfills [1] (self-reported)

## How it works
1. Inbound agents handle patient calls and inquiries around the clock.
2. Outbound campaigns notify and reschedule patients when schedules change.
3. Workflows tailored to the practice's orthopedic scheduling rules and SOPs.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Sources
1. Assort Health, "Boston Bone & Joint Institute Reduces Call Abandonment and Expands Access" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/boston-bone-joint-patient-access>
   > "With a 9% call abandonment rate, patients who couldn't get through were simply leaving."

Canonical: https://agentcasefile.com/cases/boston-bone-joint-assort-inbound-outbound-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Bloomerang using an AI agent for support?

> Bloomerang uses Fin on all support conversations. Intercom reports Fin resolving 84% automatically, with an 85% Customer Experience Score.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Bloomerang (nonprofit donor management software company)
- **Industry:** Software, Internet & Media — Nonprofit SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- automation rate: 84% [1] (self-reported)
- CX Score: 85% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "How Fin transformed support at Bloomerang" (undated; accessed 2026-10-17) <https://fin.ai/customers/bloomerang>
   > "Fast forward to today and Fin is involved in 100% of Bloomerang’s support conversations, resolving 84% of them automatically."

Canonical: https://agentcasefile.com/cases/bloomerang-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Barrington Orthopedic Specialists cut hold times with an AI scheduling agent?

> Barrington Orthopedic Specialists (26 providers, four Chicago-area sites) integrated an Assort Health voice agent into its EMR to answer scheduling calls. The vendor reports an 83% reduction in hold time, waits cut from 30+ minutes to under 5, and $120,000 a year in after-hours appointment revenue.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Barrington Orthopedic Specialists (orthopedic practice)
- **Industry:** Hospitals & Health Systems — Orthopedics
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI scheduling agent (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- hold time: 83% reduction; 30+ min to <5 min [1] (self-reported)
- after-hours revenue: $120,000 annually [1] (self-reported)
- after-hours bookings: 36 appointments booked monthly [1] (self-reported)

## How it works
1. Voice agent answers all incoming scheduling calls.
2. Follows the practice's scheduling rules to match provider, location, and appointment type.
3. Books appointments after hours.

## Risks & compliance
- Mis-scheduling or mis-routing of patients with clinical urgency; escalation rules required
- PHI handled in calls (HIPAA); EHR write access
- Vendor-published results; not independently audited

## MedTech angle
Orthopedic practices are core accounts for ortho device reps; faster patient access upstream means faster time to surgical consults and procedures.

## Sources
1. Assort Health, "Barrington Orthopedic Specialists Cuts Hold Times and Generates $120K Annually" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/barrington-orthopedic-specialists>
   > "Now the phone gets answered right away, patients get scheduled accurately, and our staff finally has the time to focus on more complex care needs."

Canonical: https://agentcasefile.com/cases/barrington-orthopedic-specialists-assort-ai-scheduling/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Austin Regional Clinic automate patient registration?

> Austin Regional Clinic (33 locations, Texas) deployed Notable intelligent intake to automate registration and reduce inaccurate registration data that drives denials. Notable reports 78% touchless registrations.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Austin Regional Clinic (physician-owned multispecialty clinic)
- **Industry:** Hospitals & Health Systems — Multispecialty clinic
- **Roles:** Patient Access & Outreach, Revenue Cycle & Prior Auth
- **Tools:** Intelligent Intake (Notable)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- touchless registrations: 78% [1] (self-reported)

## How it works
1. Patients complete digital registration before visits.
2. Automation enters data into EHR work queues the way staff would.
3. Aims to reduce registration errors and denials.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Austin Regional Clinic improves revenue cycle management by decreasing denials" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/austin-regional-clinic-powers-78-touchless-registrations>
   > "How the multispeciality clinic deployed Notable to redefine the registration process and vastly reduce inaccurate patient registration data."

Canonical: https://agentcasefile.com/cases/austin-regional-clinic-notable-touchless-registration/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Augusta Health automate pre-visit registration on MEDITECH?

> Augusta Health (Virginia; MEDITECH EHR) uses Notable automated digital intake across four hospital service lines. Notable reports 17,000 annualized digital intakes, about 7 minutes of staff time saved per intake, and 60% of encounters registered fully digitally.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Augusta Health (community hospital)
- **Industry:** Hospitals & Health Systems — Community hospital (MEDITECH)
- **Roles:** Patient Access & Outreach
- **Tools:** Automated digital intake (Notable); MEDITECH EHR (MEDITECH)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- digital intakes: 17,000 annualized [1] (self-reported)
- staff time per intake: an estimated 7 minutes saved [1] (self-reported)
- fully digital registration: 60% of all encounters [1] (self-reported)

## How it works
1. Patients receive a personalized pre-registration link by text or email.
2. They confirm demographics, upload insurance cards, and sign consents.
3. Data flows into the EHR without staff calls.

## Risks & compliance
- Automation writes into the EHR; errors propagate to registration and billing
- PHI handling (HIPAA)
- Vendor-published results; not independently audited

## Sources
1. Notable, "Transforming patient and staff engagement" (undated; accessed 2026-10-17) <https://www.notablehealth.com/customer-stories/transforming-patient-and-staff-engagement>
   > "Each digital intake saves staff an estimated 7 minutes."

Canonical: https://agentcasefile.com/cases/augusta-health-notable-digital-intake/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is AppFolio using an AI agent for customer support?

> AppFolio (20,000+ customers) uses Fin for support. Intercom reports ~65% Fin resolution and ~70% involvement.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** AppFolio (property management software company)
- **Industry:** Software, Internet & Media — Property management SaaS
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- Fin resolution rate: ~65% [1] (self-reported)
- Fin involvement rate: ~70% [1] (self-reported)

## How it works
1. Fin answers inbound support conversations using the company's help content.
2. Hands unresolved conversations to human agents.

## Risks & compliance
- Incorrect answers to customers; human review and escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Intercom (Fin), "AppFolio and Fin" (undated; accessed 2026-10-17) <https://fin.ai/customers/appfolio>
   > "We have about 20,000 plus customers, which have millions of tenants and millions of residents actually living in those properties."

Canonical: https://agentcasefile.com/cases/appfolio-fin-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did Anthropic use an AI agent for its own customer support?

> Anthropic chose to buy rather than build, deploying Intercom's Fin (powered by Claude) for customer support. Within weeks Fin reached a 50.8% resolution rate, up from 36% at launch, and saved over 1,700 hours of team time in the first month.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Anthropic (AI company)
- **Industry:** Software, Internet & Media — AI research & products
- **Roles:** Customer Support
- **Tools:** Fin AI Agent (Intercom)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- resolution rate: 50.8%, up from 36% at launch [1] (self-reported)
- time saved: over 1,700 hours in the first month [1] (self-reported)

## How it works
1. Fin handles the bulk of repetitive queries from free users over email and Messenger.
2. The team adds content snippets to raise resolution (doubling snippets lifted resolution 5.5% in a week, per source).
3. Complex queries go to human agents.

## Risks & compliance
- Vendor relationship: Fin runs on Anthropic's models

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Intercom, "Build vs buy: Why Anthropic chose Fin" (undated; accessed 2026-10-17) <https://www.intercom.com/customers/anthropic>
   > "50.8% resolution rate: More than half of all customer queries are now resolved by AI, up from 36% at launch."

Canonical: https://agentcasefile.com/cases/anthropic-fin-customer-support/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did a primary care practice use AI to cut hold times?

> Annapolis Internal Medicine (Maryland) uses an Assort Health voice agent for scheduling calls. Assort reports a 220% increase in labor capacity and a 75% reduction in hold times.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Annapolis Internal Medicine (primary care practice)
- **Industry:** Hospitals & Health Systems — Primary care
- **Roles:** Patient Access & Outreach
- **Tools:** Voice AI agent (Assort Health)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- labor capacity: 220% increase [1] (self-reported)
- hold times: 75% reduction [1] (self-reported)

## How it works
1. Voice agent answers calls and books appointments, including flu shots.
2. Staff handle complex requests.

## Risks & compliance
- PHI handling (HIPAA); EHR write access
- Escalation rules needed for clinically urgent requests
- Vendor-published results; not independently audited

## Sources
1. Assort Health, "Annapolis Internal Medicine customer story" (undated; accessed 2026-10-17) <https://www.assorthealth.com/customers/annapolis-internal-medicine-customer-story>
   > "8% of inbound calls were missed, impacting appointment volume"

Canonical: https://agentcasefile.com/cases/annapolis-internal-medicine-assort-voice-ai/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did AngelList speed a Redshift-to-Snowflake migration with Devin?

> AngelList used Devin to finish a Redshift-to-Snowflake migration, including the Metabase cutover. Cognition reports the cutover landed 5.2x faster than the team's trajectory, which had been trending six months late and would have required three or four more engineers.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** AngelList (venture fund administration platform)
- **Industry:** Banking & Financial Services — Fund administration
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- migration speed: 5.2x faster [1] (self-reported)

## How it works
1. Devin migrates queries and dashboards.
2. A data engineer owns the cutover and reviews Devin's work.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "How AngelList Completed a Redshift-to-Snowflake Migration 5.2x Faster with Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/angellist>
   > "How AngelList Completed a Redshift-to-Snowflake Migration 5.2x Faster with Devin"

Canonical: https://agentcasefile.com/cases/angellist-devin-snowflake-migration/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Airtable using AI agents for customer support?

> Airtable launched a Sierra agent on chat within four weeks and on email seven days later. Sierra reports an 80% AI resolution rate.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Airtable (software company)
- **Industry:** Software, Internet & Media — SaaS
- **Roles:** Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- AI resolution rate: 80% [1] (self-reported)

## How it works
1. AI agent resolves support requests on chat and email.
2. Escalates to support staff when needed.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited

## Sources
1. Sierra, "Airtable and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/airtable>
   > "Airtable's AI resolution rate hit 80%, a significant jump from its previous platform."

Canonical: https://agentcasefile.com/cases/airtable-sierra-support-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Airbyte use Devin to triage and fix engineering issues?

> Airbyte runs Devin on triage-and-fix workflows. Cognition reports that over 90 days Devin triaged ~990 issues, drafted 413 fixes, and ran 467 verification sessions; Airbyte merges roughly 50-60% of the resulting PRs.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** Airbyte (open-source data integration company)
- **Industry:** Software, Internet & Media — Data infrastructure (open source)
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- issues triaged (90 days): ~990 [1] (self-reported)
- fixes drafted: 413 [1] (self-reported)
- PR merge rate: roughly 50–60% [1] (self-reported)

## How it works
1. Devin triages incoming issues and drafts fixes.
2. Runs verification sessions; humans review and merge.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "Airbyte and Devin" (undated; accessed 2026-10-17) <https://devin.ai/customers/airbyte>
   > "Over the last 90 days, Devin triaged ~990 issues, drafted 413 fixes, and ran 467 verification sessions."

Canonical: https://agentcasefile.com/cases/airbyte-devin-issue-triage-fixes/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did The Access Group increase engineering output with Devin?

> The Access Group (250 products, 2,000+ engineers) adopted Devin. Cognition reports 5x more merged pull requests than a year earlier; in one test, ten items estimated at two days each came back in twenty minutes.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** The Access Group (business software company)
- **Industry:** Software, Internet & Media — Business management software (UK)
- **Roles:** Software Engineering
- **Tools:** Devin (Cognition)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- merged pull requests: 5x more than a year ago [1] (self-reported)

## How it works
1. Engineers delegate backlog items to Devin.
2. Merged pull requests tracked as output measure.

## Risks & compliance
- Agent-written code requires human review and testing
- Vendor-published results; not independently audited

## Sources
1. Cognition (Devin), "How The Access Group 5x'd engineering output across 250 products and 2,000+ engineers" (undated; accessed 2026-10-17) <https://devin.ai/customers/access-group>
   > "They gave those ten items to Devin: all ten came back in twenty minutes."

Canonical: https://agentcasefile.com/cases/access-group-devin-engineering-output/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did 1-800-Flowers pilot AI agents for customer service?

> 1-800-Flowers.com piloted Decagon AI agents for customer service. Decagon reports the pilot closed with CSAT nearing 93%, double-digit points higher than previous automation.

- **Evidence grade:** B — Named organization with a quantified outcome, from the vendor or the organization's own announcement.
- **Organization:** 1-800-Flowers.com (gifting retailer)
- **Industry:** Consumer Products & Retail — E-commerce / gifting
- **Roles:** Customer Support
- **Tools:** Decagon AI agent (Decagon)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- pilot CSAT: nearing 93% [1] (self-reported)

## How it works
1. Decagon AI agent answers inbound customer conversations.
2. Escalates to human agents when needed.

## Risks & compliance
- Incorrect answers to customers; escalation paths needed
- Vendor-published results; not independently audited

## Sources
1. Decagon, "1-800-Flowers.com and Decagon" (undated; accessed 2026-10-17) <https://decagon.ai/case-studies/1-800-flowers>
   > "The pilot closed with CSAT nearing 93% - double-digit points higher than previous automation."

Canonical: https://agentcasefile.com/cases/1-800-flowers-decagon-ai-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is Sanofi using an AI agent to prepare field reps for physician visits?

> Sanofi launched Concierge for Field, an AI agent built with Snowflake Cortex AI that prepares global sales reps for physician visits: pre-call plans, highest-priority physicians, and past engagement. Snowflake said it reduces hours of manual research to seconds; no measured outcome was published.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Sanofi (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Big Pharma
- **Roles:** Field Sales Reps
- **Tools:** Concierge for Field (Sanofi (built on Snowflake Cortex AI))
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Rep asks for a pre-call plan in a single conversation.
2. Agent surfaces highest-priority physicians by specialty and prescribing history and reviews past engagement.
3. Full plan emailed to the rep.

## Risks & compliance
- Use of prescribing data must follow privacy and market rules
- Vendor time-savings claim ("hours to seconds") is qualitative

## MedTech angle
Pre-call planning agents of this kind are the closest pharma analog to device-rep territory and case prep tools.

## Sources
1. Fierce Pharma, "Sanofi taps Snowflake for AI field agent help" (2026-06-02) <https://www.fiercepharma.com/marketing/sanofi-taps-snowflake-ai-field-agent-help>
   > "The agent, called Concierge for Field, was built with Snowflake Cortex AI."

Canonical: https://agentcasefile.com/cases/sanofi-snowflake-concierge-for-field-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is KARL STORZ using AI agents in medical device sales?

> KARL STORZ, the German endoscopy maker, is rolling out Salesforce Agentforce Sales and CPQ globally and presented its experience with AI agents for its sales and service teams at MedtecLIVE 2026. No quantified outcomes were published in the session listing.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** KARL STORZ (medical device manufacturer)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Surgical endoscopy & imaging
- **Roles:** Field Sales Reps, Sales Ops & Quoting, Customer Support
- **Tools:** Agentforce Sales (Salesforce); Salesforce CPQ (Salesforce)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Global rollout of Agentforce Sales and CPQ (configure-price-quote).
2. AI agents aim to increase sales efficiency and relieve sales staff of routine work.
3. Next phase per the session: customer service and the customer portal.

## Risks & compliance
- Pricing/quote accuracy for regulated products
- Local-market requirements vs. global standards (named in the source)

## MedTech angle
One of the first named European MedTech companies to publicly describe agentic AI for its sales force; session held in German at MedtecLIVE Nuremberg, 6 May 2026.

## Notes
- Source is a conference session description co-presented with Salesforce; treat claims as self-described.

## Sources
1. MedtecLIVE, "How KARL STORZ Empowers its Sales and Service Teams with AI Agents" (2026-05-06) <https://www.medteclive.com/en/e/how-karl-storz-empowers-its-sales-and-service-teams-with-ai-agents.79721>
   > "Transforms its digital landscape through the global rollout of Agentforce Sales and CPQ,"

Canonical: https://agentcasefile.com/cases/karl-storz-agentforce-sales-cpq/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Which pharma companies are adopting Veeva's AI agents for field reps?

> Novo Nordisk is adopting Veeva Vault CRM with Veeva AI Agents such as Pre-call Agent (call prep) and Voice Agent (spoken call notes), which became available in December 2025. Novo's field systems director says the agents 'will drive efficiencies'; no measured outcomes have been published yet.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Novo Nordisk (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Pharmaceuticals
- **Roles:** Field Sales Reps
- **Tools:** Veeva AI Agents (Pre-call, Voice, Free Text) (Veeva Systems); Vault CRM (Veeva Systems)
- **Status:** announced · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Pre-call Agent gathers insights and suggested actions from data, content, activity, and trends before a rep's call.
2. Voice Agent lets the rep capture call information and follow-ups by speaking.
3. Free Text Agent flags potential accuracy or compliance issues in call notes.

## Risks & compliance
- Call notes must not contain patient identifiers (Free Text Agent screens for this)
- Promotional compliance

## MedTech angle
The same pre-call / voice-note / compliance-screen pattern applies directly to device reps' call reporting.

## Sources
1. Veeva Systems, "Veeva AI Agents Now Available to Increase Productivity and Customer Centricity" (2025-12-03) <https://www.veeva.com/resources/veeva-ai-agents-now-available-to-increase-productivity-and-customer-centricity/>
   > "Vault CRM and Veeva AI Agents like Pre-call Agent and Voice Agent will drive efficiencies and allow the field to focus on the value parts of their jobs."

Canonical: https://agentcasefile.com/cases/novo-nordisk-veeva-precall-voice-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can an AI agent speed up pharma promotional review (MLR)?

> Moderna is an early-access user of Veeva AI's Quick Check Agent in PromoMats, which scans promotional content against editorial, brand, market, channel, and compliance guidelines before medical-legal-regulatory (MLR) review. Moderna says it moves the company closer to parts of MLR becoming 'nearly touch-free'; no metrics were published.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Moderna (biotech company)
- **Industry:** Pharma, Biotech & Life Sciences — Biotech / vaccines
- **Roles:** Marketing Ops & MLR
- **Tools:** Quick Check Agent (PromoMats) (Veeva Systems)
- **Status:** pilot · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Marketer submits a draft piece into PromoMats.
2. Quick Check Agent scans it against editorial, brand, market, channel, and compliance guidelines.
3. Issues are surfaced before the piece enters formal MLR review.

## Risks & compliance
- Final MLR approval remains human
- Off-label or unsubstantiated claims must still be caught by reviewers

## MedTech angle
Device marketing teams run the same promotional-review gauntlet; pre-review checks are a low-risk first agent.

## Sources
1. Veeva Systems, "Veeva AI Agents Now Available to Increase Productivity and Customer Centricity" (2025-12-03) <https://www.veeva.com/resources/veeva-ai-agents-now-available-to-increase-productivity-and-customer-centricity/>
   > "What we’ve learned as an early access user is that the Veeva AI Quick Check Agent moves Moderna closer to a process where parts of MLR could become nearly touch-free."

Canonical: https://agentcasefile.com/cases/moderna-veeva-quick-check-agent-mlr/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How is Takeda using AI agents with its pharma field teams?

> Takeda is rolling out Agentforce on Salesforce Life Sciences Cloud and Data 360 so field reps arrive armed with an HCP's recent research, email opens, and questions, and so leaders can see trends across engagements. Takeda describes this as an aspiration and early rollout; no outcomes are reported.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Takeda Pharmaceutical (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Pharmaceuticals
- **Roles:** Field Sales Reps, Commercial Leadership
- **Tools:** Agentforce (Salesforce); Life Sciences Cloud (Salesforce); Data 360 (Salesforce); MuleSoft (Salesforce)
- **Status:** pilot · **Agent pattern:** multi-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Data 360 and MuleSoft pull HCP interactions from rep visits, email, and WhatsApp into one view.
2. Agents structure unstructured data into actionable insights for reps before a visit.
3. Leadership agents summarize what interactions are happening and what resonates across hospitals.
4. 'Utility agents' are built in Agentforce or surfaced from other agents via Model Context Protocol.

## Risks & compliance
- Least-privilege data access per agent (stressed by Takeda)
- HCP consent for channel data

## MedTech angle
Multiple reps calling on the same hospital is common in device sales too; the leadership-insights agent addresses that overlap.

## Notes
- Takeda states reps spent about a day each week on administrative tasks; the agent goal is to reduce that, but no result is reported.

## Sources
1. Salesforce, "Providers suffer from data overload. Agentforce powers Takeda's digital solution for healthcare engagement." (2025-10-13) <https://www.salesforce.com/customer-stories/takeda-garth-hemmingsen-interview/>
   > "Agentforce gives leadership deep insights at their fingertips — what interactions are happening, what’s resonating, and what the needs are in that environment."

Canonical: https://agentcasefile.com/cases/takeda-agentforce-hcp-engagement-insights/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can a small biotech use an AI SDR agent to qualify conference leads?

> AUM BioTech's CEO, with no prior coding experience, built lead-gen forms and a support chatbot on Salesforce and is finishing an Agentforce SDR agent to respond to thousands of conference leads, score them, book meetings, and hand top leads to human reps. The SDR agent was not yet live at publication and no outcomes are reported.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** AUM BioTech (small biotech company)
- **Industry:** Pharma, Biotech & Life Sciences — Biotech research products
- **Roles:** Sales Development (SDR), Customer Support
- **Tools:** Agentforce (Salesforce); Marketing Cloud (Salesforce)
- **Status:** announced · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Leads are collected at 5–6 scientific conferences a year (up to 10,000 leads).
2. SDR agent responds instantly, scores each lead, and answers basic questions in multiple languages, 24/7.
3. Books meetings on request; higher-priority leads go directly to human sales reps.
4. Marketing Cloud campaigns nurture the rest.

## Risks & compliance
- Guardrails restricting agent access to sensitive data (stated by the CEO)
- Email consent for outreach

## MedTech angle
Trade-show lead follow-up is a common pain for small device and diagnostics companies.

## Sources
1. Salesforce, "How a Small BioTech Firm is Using Agentic AI to Take on Industry Giants" (2025-09-18) <https://www.salesforce.com/news/stories/aum-biotech-agentforce-sdr-agent/>
   > "Now, he’s putting the finishing touches on a Sales Development Representative (SDR) agent, which will allow the company to instantly respond to thousands of leads, score each one, and identify the best candidates for more extensive outreach."

Canonical: https://agentcasefile.com/cases/aum-biotech-sdr-agent/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Which health systems have deployed AI agents for nurses and post-op calls?

> Hackensack Meridian Health has developed a nurse AI agent that sifts large amounts of data and offers next-best-action recommendations, plus an agent that assists with post-operative calls to patients, built with Google Cloud. Fierce Healthcare reported the deployment in April 2025 without outcome figures.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Hackensack Meridian Health (health system)
- **Industry:** Hospitals & Health Systems — Health system (18 hospitals)
- **Roles:** Clinicians & Clinical Ops, Patient Access & Outreach
- **Tools:** Gemini / Vertex AI (Google Cloud)
- **Status:** production · **Agent pattern:** copilot · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Gemini-based app summarizes medical records via Vertex AI Search.
2. Nurse agent reviews large volumes of patient data and suggests next best actions.
3. Post-op agent assists with follow-up calls to patients after surgery.

## Risks & compliance
- Clinical decision support; nurses retain judgment
- PHI handling

## MedTech angle
Post-operative follow-up after orthopedic and surgical cases is a touchpoint device companies care about.

## Sources
1. Fierce Healthcare, "Google Cloud unveils tools to build multi-agent AI systems as Highmark, Hackensack Meridian forge ahead on AI innovation" (2025-04-09) <https://www.fiercehealthcare.com/ai-and-machine-learning/google-cloud-builds-out-ai-agent-capabilities-healthcare-highmark-health>
   > "Hackensack has developed a nurse AI agent to help nurses sift through large amounts of data fairly quickly and offer next best action recommendations and an agent to assist with post-operative calls to patients."

Canonical: https://agentcasefile.com/cases/hackensack-meridian-nurse-postop-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How does Toyota of Orlando use AI to route sales leads?

> Toyota of Orlando receives 200+ daily leads in inconsistent formats. Its operations director built a 38-step Zap that uses AI to extract, clean, and route leads, after first building a stopgap CRM in Zapier during the month-long CDK ransomware outage. The Zapier story does not attribute a measured outcome to the AI routing.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Toyota of Orlando (auto dealership)
- **Industry:** Consumer Products & Retail — Automotive retail (dealership)
- **Roles:** Sales Ops & Quoting, Sales Development (SDR)
- **Tools:** Zapier (AI Zap workflows) (Zapier)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. AI extracts and cleans lead data from multiple sources.
2. Routes leads to the right salespeople.
3. Served as a fallback CRM during an outage.

## Risks & compliance
- Data quality in AI-extracted lead fields
- Vendor-published

## Sources
1. Zapier, "How Toyota of Orlando uses AI and automation to keep sales running—through outages and beyond" (undated; accessed 2026-10-17) <https://zapier.com/customer-stories/toyota-orlando>
   > "Toyota of Orlando receives 200+ daily leads in inconsistent formats, making lead routing error-prone and time-consuming."

Canonical: https://agentcasefile.com/cases/toyota-of-orlando-zapier-ai-lead-routing/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is Sutter Health using AI agents for patient engagement?

> Sutter Health (25 hospitals, 3.5 million patients) is working with Sierra on AI agents for patient engagement. The Sierra page does not publish quantified outcomes.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Sutter Health (health system)
- **Industry:** Hospitals & Health Systems — Integrated health system
- **Roles:** Patient Access & Outreach, Customer Support
- **Tools:** Sierra AI agent (Sierra)
- **Status:** production · **Agent pattern:** autonomous-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. AI agents engage patients on Sutter's behalf.
2. Described by Sierra as a blueprint for health systems adopting AI in patient engagement.

## Risks & compliance
- Wrong or hallucinated answers to customers; needs guardrails and handoff
- Vendor-published results; not independently audited
- PHI handling (HIPAA)

## Sources
1. Sierra, "Sutter Health and Sierra" (undated; accessed 2026-10-17) <https://sierra.ai/customers/sutter-health>
   > "With 25 hospitals and over 3.5 million patients across Northern California, Sutter's experience offers a practical blueprint for health systems ready to bring AI into patient engagement."

Canonical: https://agentcasefile.com/cases/sutter-health-sierra-patient-engagement/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is Boehringer Ingelheim using AI agents in MLR content review?

> Boehringer Ingelheim joined the Veeva AI Early Adopter Program and spent four months piloting Quick Check Agent and Content Agent in PromoMats for medical, legal, and regulatory (MLR) review. Veeva's story shares lessons but no quantified outcomes.

- **Evidence grade:** C — Named organization, no specific outcome (deployment announced only, or only vague qualitative claims).
- **Organization:** Boehringer Ingelheim (pharmaceutical company)
- **Industry:** Pharma, Biotech & Life Sciences — Big Pharma
- **Roles:** Marketing Ops & MLR
- **Tools:** Quick Check Agent and Content Agent (Veeva AI for PromoMats) (Veeva)
- **Status:** pilot · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Quick Check Agent pre-screens promotional content before MLR submission.
2. Content Agent assists with content work inside PromoMats.
3. Piloted with Veeva Business Consulting within live processes.

## Risks & compliance
- Promotional compliance; AI checks do not replace MLR reviewers
- Vendor-published

## Sources
1. Veeva, "Boehringer Ingelheim Is a Veeva AI Early Adopter" (undated; accessed 2026-10-17) <https://www.veeva.com/customer-stories/beyond-the-ai-hype-learnings-implementing-vault-ai-in-promomats/>
   > "Over four months, Boehringer partnered closely with Veeva Business Consulting to pilot Veeva AI capabilities — including Quick Check Agent and Content Agent"

Canonical: https://agentcasefile.com/cases/boehringer-ingelheim-veeva-ai-promomats-pilot/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Is there an AI assistant for med device case scheduling and field inventory?

> Movemedical, a field inventory platform for medical device companies, released Ask Move AI in its June 2026 platform release: a mobile assistant that lets reps create cases by voice or text, look up case inventory, reschedule cases, and look up invoices. It is opt-in for existing customers; no named customer deployment or outcome is reported yet.

- **Evidence grade:** D — Anonymous or community report, or a demo/concept/product launch with no named deployment. Shown, but labeled and excluded from "proven" filters.
- **Organization:** Movemedical (vendor launch) (MedTech software vendor)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Field inventory & case management
- **Roles:** Field Sales Reps, Field Service & Case Coverage
- **Tools:** Ask Move AI (Movemedical)
- **Status:** launch · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- None quantified in sources.

## How it works
1. Rep asks by voice or text on mobile.
2. Agent layer executes the request on Movemedical's existing deterministic case/inventory platform.
3. Every interaction is logged, audited, and authenticated by the platform.
4. Roadmap items per the source: usage capture, data insights, consignment optimization, demand forecasting.

## Risks & compliance
- Case data may contain PHI (platform states SOC 2, HIPAA, HITRUST certification)
- Customer data not used to train models (per vendor)

## MedTech angle
Targets the daily loaner-kit and case-coverage workflow that orthopedic and spine reps live in.

## Sources
1. PR Newswire (Movemedical), "Movemedical Launches Ask Move AI, the First AI Assistant Purpose-Built for Med Device Field Operations" (2026-07-07) <https://www.prnewswire.com/news-releases/movemedical-launches-ask-move-ai-the-first-ai-assistant-purpose-built-for-med-device-field-operations-302817222.html>
   > "Ask Move AI launches with four capabilities available immediately on mobile: smart case creation by voice or text, case inventory lookup, case rescheduling, and invoice lookup."

Canonical: https://agentcasefile.com/cases/movemedical-ask-move-ai-case-workflow/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# Can AI agents build territory plans for medical device reps?

> MedScout launched 'Strategies', AI agents that analyze referral networks, procedure volumes, and payer dynamics to deliver prioritized, ready-to-work territory plans to MedTech reps; it says the agents were available to customers from late 2025. The deployment examples in the announcement are anonymous (e.g., a surgical robotics company), so this case is graded D.

- **Evidence grade:** D — Anonymous or community report, or a demo/concept/product launch with no named deployment. Shown, but labeled and excluded from "proven" filters.
- **Organization:** Unnamed MedTech customers of MedScout (medical device companies)
- **Industry:** Medical Devices & Diagnostics (MedTech) — MedTech commercial software
- **Roles:** Field Sales Reps, Sales Ops & Quoting, Commercial Leadership
- **Tools:** MedScout Strategies (MedScout)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- agent adoption (vendor-wide): customers deployed 81% more unique AI agents in January than the month before [1] (self-reported)

## How it works
1. Each 'Strategy' agent is configured around a company's commercial context and definition of a best-fit target.
2. It analyzes market and account-level referral networks, procedure volumes, and payer dynamics.
3. Reps receive a prioritized, territory-specific list of providers and facilities to work.
4. Example from the source: identifying hospitals with falling TAVR volume but high cath-lab capacity for a cardiology device company.

## Risks & compliance
- Use of claims-derived physician data
- Targeting must stay within on-label indications

## MedTech angle
Core rep workflow: turning commercial strategy into a ranked call list. GE HealthCare's Surgical Innovations CMO is quoted as a MedScout partner, but the source does not say which features GE uses.

## Notes
- The GE HealthCare quote in the release describes MedScout as a strategic partner; we do not attribute the Strategies agents to GE because the source does not.

## Sources
1. MedScout, "MedScout Raises $10M and Launches AI Agents for MedTech Commercial Teams" (2026-02-24) <https://www.medscout.io/resources/medscout-raises-10m-and-launches-ai-agents-for-medtech-commercial-teams>
   > "In January, our customers deployed 81% more unique AI agents than the month before"

Canonical: https://agentcasefile.com/cases/medscout-strategies-territory-agents/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How do specialty pharmacies use voice AI agents for benefit verification calls?

> Infinitus says its voice AI agents completed over 60,000 benefit-verification and prior-authorization follow-up tasks for an unnamed leading specialty pharmacy in the first half of 2025, with each verification call freeing an average of 29 minutes of staff time. The pharmacy is not named, so this is graded D.

- **Evidence grade:** D — Anonymous or community report, or a demo/concept/product launch with no named deployment. Shown, but labeled and excluded from "proven" filters.
- **Organization:** Unnamed leading specialty pharmacy (specialty pharmacy)
- **Industry:** Pharma, Biotech & Life Sciences — Specialty pharmacy
- **Roles:** Patient Access & Outreach, Revenue Cycle & Prior Auth
- **Tools:** Infinitus voice AI agents (Infinitus Systems)
- **Status:** production · **Agent pattern:** voice-agent · **Autonomy:** 3/4
- **Last verified:** 2026-10-17

## Reported outcomes
- tasks completed: over 60,000 tasks in H1 2025 [1] (self-reported)
- staff time: average of 29 minutes unlocked per benefit verification call [1] (self-reported)

## How it works
1. Agent places outbound calls to payers and navigates IVRs and hold queues.
2. Collects benefits, eligibility, and prior-authorization status.
3. Returns structured results to pharmacy staff; exceptions go to humans.

## Risks & compliance
- PHI on every call
- Accuracy of captured benefit details

## MedTech angle
Durable medical equipment and device reimbursement teams make the same payer calls.

## Sources
1. Infinitus Systems, "Infinitus and IBM Consulting partner to bring agentic AI into healthcare entities at scale" (2025-09-18) <https://www.infinitus.ai/blog/infinitus-and-ibm-consulting-partner-to-bring-agentic-ai-into-healthcare-entities-at-scale/>
   > "Infinitus performed over 60,000 such tasks for the specialty pharmacy during the first half of 2025 alone"

Canonical: https://agentcasefile.com/cases/specialty-pharmacy-infinitus-benefit-verification/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How did a MedTech diagnostics company use Agentforce for field service?

> An unnamed MedTech company that services diagnostic equipment in hospitals and labs combined Salesforce Agentforce and Field Service via partner Redtag. Redtag reports 100% of service visits auto-scheduled, case resolution time cut by 50%, and a 30% improvement in technician availability.

- **Evidence grade:** D — Anonymous or community report, or a demo/concept/product launch with no named deployment. Shown, but labeled and excluded from "proven" filters.
- **Organization:** Unnamed MedTech diagnostics equipment company (medical device / diagnostics company)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Diagnostic equipment service
- **Roles:** Field Service & Case Coverage
- **Tools:** Agentforce + Field Service (Salesforce); Implementation (Redtag)
- **Status:** production · **Agent pattern:** workflow-agent · **Autonomy:** 2/4
- **Last verified:** 2026-10-17

## Reported outcomes
- auto-scheduled visits: 100% [1] (self-reported)
- case resolution time: cut by 50% [1] (self-reported)
- technician availability: 30% improvement [1] (self-reported)

## How it works
1. Agentforce pre-screens service cases.
2. Visits auto-scheduled and technicians routed in Field Service.

## Risks & compliance
- Equipment downtime affects patient care
- Anonymous; partner-published

## MedTech angle
Field service and case coverage are core device-company operations.

## Sources
1. Redtag, "Cutting Downtime by 50%, and Streamlining Service Delivery in MedTech" (undated; accessed 2026-10-17) <https://www.redtag.pro/our-stories/cutting-downtime-by-50-and-streamlining-service-delivery-in-medtech>
   > "Case resolution time cut by 50%"

Canonical: https://agentcasefile.com/cases/medtech-diagnostics-redtag-agentforce-field-service/ · License: CC BY 4.0 (summary) · Source: Agent Casefile

---

# How can device and lab reps log visit notes by voice with an AI agent?

> Salesforce partner Demand Chain built an Agentforce assistant for an unnamed global lab-services and medical technology company: field reps dictate visit summaries on Salesforce Mobile and the agent turns speech into structured trip-report records. The partner reports end-of-day admin backlogs were eliminated, but the customer is anonymous and no numbers are given.

- **Evidence grade:** D — Anonymous or community report, or a demo/concept/product launch with no named deployment. Shown, but labeled and excluded from "proven" filters.
- **Organization:** Unnamed global lab services & medical technology company (diagnostics / medtech)
- **Industry:** Medical Devices & Diagnostics (MedTech) — Lab services & diagnostics
- **Roles:** Field Sales Reps, Commercial Leadership
- **Tools:** Agentforce (Salesforce); Salesforce Mobile App (Salesforce)
- **Status:** pilot · **Agent pattern:** workflow-agent · **Autonomy:** 1/4
- **Last verified:** 2026-10-17

## Reported outcomes
- admin backlog: reps log visits instantly; end-of-day admin backlogs eliminated (qualitative) [1] (self-reported)

## How it works
1. Rep opens the assistant in the Salesforce mobile app and dictates notes after a visit.
2. Agent extracts structured fields such as provider name and visit type.
3. Spoken notes are converted to bullet-pointed insights on the existing Trip Report record.
4. Progressive prompts and error handling keep the rep on track.

## Risks & compliance
- Dictated notes can capture patient identifiers; needs PHI screening
- Accuracy of structured extraction should be reviewed by the rep

## MedTech angle
Same pattern as pharma 'voice call reporting' agents, applied to a lab/MedTech field force.

## Notes
- Source page is undated; accessed 2026-10-17.

## Sources
1. Demand Chain, "How a Global Lab & Med Tech Leader Boosted Field Sales Productivity with an AI Assistant" (undated; accessed 2026-10-17) <https://www.demandchain.com/ai-assistant-field-sales-voice-to-text/>
   > "Reps log visits instantly with hands-free voice input, eliminating end-of-day admin backlogs."

Canonical: https://agentcasefile.com/cases/lab-medtech-field-visit-notes-agentforce/ · License: CC BY 4.0 (summary) · Source: Agent Casefile