Episode 24 · Healthcare Innovation · August 4, 2026
AI At The Bedside: What Clinicians Actually Notice
Dr. Amara Osei, MD — Chief Medical Information Officer, Northline Health System
Executive Summary
Dr. Osei led one of the largest ambient AI documentation deployments in the country. In this conversation she separates vendor narrative from floor reality: adoption curves collapse without a clinical champion on every unit, note quality gains show up before time savings do, and the biggest risk isn't hallucination — it's clinicians quietly disengaging from their own documentation.
Key Takeaways
- 01Time savings are the wrong first metric — measure cognitive load and note completeness in the first 90 days.
- 02Every unit needs a clinician champion; centralized rollouts stall at roughly 40% adoption.
- 03Ambient tools change what clinicians say out loud in the room, which changes the patient relationship.
- 04Governance must define who owns an AI-drafted note before go-live, not after the first incident.
- 05The best predictor of sustained use is whether the tool survives a bad night shift.
YEM's Insider Take
We keep evaluating healthcare AI like it's software. It isn't. It's a workflow change wearing a software costume. The organizations winning here budgeted more for change management than for licenses — and they are the only ones still using the tool in year two.
Transcript
YEM: You said something in our prep call that stuck with me — that the pilot unit was chosen wrong on purpose. DR. OSEI: Deliberately. Everyone pilots on their highest performing unit and then wonders why the results don't generalize. We went to the unit with the worst documentation backlog in the system. If it worked there, we'd believe it.
The Insider Question