ElevenAgents for Healthcare: Agents Patients Trust
·2026.11.18 21:00
Key point
The platform integrates Scribe v3 Live Medical for real-time transcription and supports HIPAA-eligible deployments with Zero Retention Mode.
Details
- ElevenLabs introduces ElevenAgents for Healthcare, a platform enabling healthcare teams to deploy AI agents for patient and provider calls with full control over behavior and guardrails.
- Cadence, a chronic condition monitoring service, reduced median time to reach patients from nearly two hours to 3.5 minutes, with patients answering agent calls 47% more often than calls from nurses.
- Scribe v3 Live Medical, the new real-time medical transcription model, achieved the lowest word error rate on the PriMock57 benchmark, with 17% fewer errors than the next closest model tested.
- Angle Health uses the agents to resolve more than 90% of provider calls without transferring to a human agent.
Performance and Accuracy
- On the Eka dataset of over 3,600 English medical recordings, Scribe v3 Live Medical has an 11% error rate compared to 11.54% for the general Scribe v3 Live model.
- Reducing end-to-end latency by an average of 16% correlated to a 26% lift in conversation success rate in ElevenAgents deployments.
- The system supports up to 10,000 custom terms for specific plans, therapies, and locations, and handles language switching between English and Spanish.
Deployment and Safety
- HIPAA-eligible deployments are available for Enterprise customers who sign a Business Associate Agreement (BAA) and enable Zero Retention Mode, ensuring audio and transcripts are not stored on ElevenLabs infrastructure.
- Built-in and custom guardrails allow teams to define rules in plain language, such as transferring symptom questions to clinicians, and require identity confirmation before sensitive actions.
- Forward-deployed engineers assist with integrating legacy systems and custom software, while templates like Healthcare Receptionist allow teams to launch working agents in minutes.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.