The problem
Documentation burden is the single biggest driver of clinician burnout. Existing scribing tools generated text — but not the structured note clinicians could sign. Citations to the source utterance were essentially nonexistent.
Ambient capture existed as a demo; production needed citations, PHI boundaries, and an audit trail the compliance team could defend.
Our approach
We built a scribe pipeline with two non-negotiable invariants:
- Every claim has a source. The clinician can hover any sentence and see the ambient transcript span that produced it.
- Templates are owned by the clinical team. Note structure (SOAP, narrative, etc.) is defined per-specialty; the agent fills the slots.
RAG-grounded scribe agents draft clinician-ready notes from ambient audio. Human review stays in the loop for sign-off.
The first version we shipped had a 12% citation-failure rate. The third had 0.8%. The evals caught every regression.
— Eng lead, Healthtech, US
Architecture notes
Data and agent lanes: ambient capture → chunking → retrieval → draft → citation check → clinician sign-off. Interactive portal embeds (architecture diagram, sample run) stay app-local; this CMS body is the narrative.
Outcomes
| Metric | Result |
|---|---|
| Time saved / clinician / week | 6.5h (52%) |
| Citation-failure rate | 0.8% |
| Clinician sign-off rate | 94% |
| Notes with source citations | 100% |
| Escalations to human rewrite | under 8% |
For the first time in five years, I left the clinic on time. The notes were waiting for me, with the right sources.
— Attending physician, Outpatient, US