An ambient AI scribe that gives clinicians their evenings back
Client: Meridian Health Group
- 2 hrs
- saved per clinician each day
- 40%
- less time on documentation
- same-day
- note completion rate
The challenge
Clinicians across a multi-site provider group were spending two-plus hours a day writing up visit notes after hours — a leading driver of burnout and a hard ceiling on how many patients they could see.
What we built
We built an ambient documentation assistant that listens to a consented visit, drafts a structured clinical note in the provider's template, and hands it to the clinician for a fast review-and-sign — with strict data-handling and a full audit trail.
How we did it
- 01
Shadowed clinicians across three specialties to learn how notes are actually structured — the product was designed around their real workflow, not a generic template.
- 02
Engineered for privacy from day one: consented capture, PHI minimization before any model call, encryption throughout, and a complete access audit trail.
- 03
Built a domain-tuned transcription and summarization pipeline that produces structured notes in each specialty's format, with an eval set of real anonymized visits gating every change.
- 04
Kept the clinician in control: every note is a fast review-and-sign, never an auto-submit — trust was earned before autonomy was expanded.
- 05
Rolled out site by site with usage analytics, tuning the model against the feedback until adoption became voluntary.
The result
Clinicians reclaimed most of their documentation time, notes were completed the same day, and the group could expand capacity without adding administrative headcount.
Stack
Services used
- AI & GenAI
- Product Engineering
More work