Executive Summary
- Physicians spend up to 2 hours a night 'pajama time' doing data entry into EMRs (Electronic Medical Records).
- An ambient pipeline records the exam room via an Apple Watch or phone, runs secure STT, and uses an LLM to generate a perfect SOAP note.
- This specific pipeline is the highest adopted AI tool in healthcare, reducing physician burnout massively.
Documentation Time Savings
1.5 HrsDaily Per Doctor
Total elimination of manual keyboard data entry for patient charting.
1. The Edge Audio Capture
The raw audio of the doctor-patient conversation is captured via a secure mobile app. It is never stored on the local device. It is streamed instantly via WebRTC to a HIPAA-compliant VPC.
Physician Time Allocation (15 Min Appointment)
Staring at the Computer (Legacy)8
Patient Eye Contact (Legacy)7
Patient Eye Contact (Ambient AI)15
Diarization Confidence
The Speech-to-Text model must perfectly execute 'diarization' (who said what). If the patient says, 'My father had a heart attack at 50', the LLM must not assign that medical history to the patient themselves.
2. Summarization to Schema
The transcript is messy. The doctor and patient talk about the weather, kids, and then back to symptoms. The LLM extracts only the clinical facts and maps them strictly to the SOAP format (Subjective, Objective, Assessment, Plan).
3. The EMR Push
Using the Epic FHIR API, the structured JSON payload is pushed directly into the patient's chart as a 'Draft'. The doctor clicks one button to review and digitally sign at the end of the day.