UCLA Health
Verified evidenceCreator to judgeAmbient tools can draft notes in practice, but vendor-specific efficiency and error results require comparison.
Healthcare screening · California, United States
Collections: Human still decides · Regulated autonomy · Negative results
Executive brief
Treat ambient scribes as distinct products with measurable differences, not as one uniform category.
AI value · Time writing each note
VerifiedNabla arm fell from 4:30 to 3:49; a 9.5% greater reduction than control.
DAX did not significantly improve time versus control.
Before
Conducts the visit and writes the note in the EHR. → Completes and reviews documentation.
After
Records the conversation and generates a draft note. → Reviews, corrects, and signs the draft.
Human boundary
Physicians decide whether any generated content enters the record.
Why it matters
Ambient tools can draft notes in practice, but vendor-specific efficiency and error results require comparison.
Before
How the work ran before the change.
Outpatient physician
Conducts the visit and writes the note in the EHR.
ControlPhysician signature.
Physician
Completes and reviews documentation.
ControlClinical accountability.
What changed
Ambient tools can draft notes in practice, but vendor-specific efficiency and error results require comparison.
Decision rightHuman moves from creator to judge
After
How the same work runs now.
Ambient AI scribe
Records the conversation and generates a draft note.
ControlTwo-month randomized clinical deployment.
Physician
Reviews, corrects, and signs the draft.
ControlPhysician vigilance for clinically significant inaccuracies.
Exception path
Inaccurate output is corrected or discarded and documented manually; adverse events follow clinical safety reporting.
Work removed
Decision authority
Physicians decide whether any generated content enters the record.
Before
Outpatient physician
Conducts the visit and writes the note in the EHR.
Control: Physician signature.
Physician
Completes and reviews documentation.
Control: Clinical accountability.
After
Ambient AI scribe
Records the conversation and generates a draft note.
Control: Two-month randomized clinical deployment.
Physician
Reviews, corrects, and signs the draft.
Control: Physician vigilance for clinically significant inaccuracies.
Work that left the path
Human role before
Physicians composed notes directly.
Human role after
Physicians edit and validate drafts while retaining complete clinical and signature responsibility.
AI roleDecision mode: draft documentation only; no diagnosis, order, or signature authority.
Control arm fell from 4:22 to 4:04 per note.→Nabla arm fell from 4:30 to 3:49; a 9.5% greater reduction than control.
November 4, 2024-January 3, 2025. · 238 physicians, 14 specialties, about 72,000 encounters.
DAX did not significantly improve time versus control.
Usual physician-authored documentation.→One mild adverse event; clinically significant inaccuracies were reported occasionally for both products.
Two-month randomized trial. · Two scribe products across 238 physicians.
Requires continuing physician vigilance; product performance was not uniform.
Anti-pattern
Scaling a category-wide claim when one tested product did not beat control.
Questions
Portability conditions
Reputation risk
medium
Evidence and authority
Current · updated
1 peer reviewed; publication outcomes are verified.
Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 868ad97f92d5e208
The Permanente Medical Group · Creator to judge
Ambient tools can draft clinical notes, while physicians remain the editors and signers.
United Kingdom Serious Fraud Office · Creator to judge
AI can perform first-pass privilege screening at scale while lawyers retain privilege and prosecution judgments.
Tempus AI TIME / TriHealth Cancer Institute · Creator to judge
Continuous population screening can connect eligible patients to a standardized activation lane.
Google · Creator to judge
Large migrations can combine machine-found locations, LLM edits, automated checks, and engineer review.