At a glance
- Source
- Lukac, Turner, Vangala, Chin, Khalili, Shih, Sarkisian, Cheng and Mafi (2025). Ambient AI Scribes in Clinical Practice: A Randomized Trial. NEJM AI, 2(12). doi:10.1056/AIoa2501000
- How it was done
- Doctors were assigned by chance to one of two AI scribes, Microsoft's DAX Copilot or Nabla, or to carry on as usual. Time spent writing notes came from the health-record system. Randomised experiment. People were assigned by chance to work with or without the tool, and the two groups were compared.
- Who took part
- 238 outpatient doctors in 14 specialties.
- Where
- Outpatient clinics of UCLA Health in Los Angeles, United States, from 4 November 2024 to 3 January 2025.
“Nabla reduced time-in-note versus the control.”
What it found
An AI scribe records the conversation with a patient and drafts the visit note.
- How often doctors used it. Doctors used DAX Copilot in 33.5% of 24,696 visits and Nabla in 29.5% of 23,653 visits.
- Time on notes. Doctors with Nabla spent 9.5% less time writing notes than doctors without a scribe. The likely range runs from 1.8% to 17.2% less. Doctors with DAX Copilot showed no clear change (1.7% less, with a likely range from 9.4% less to 5.9% more).
- Burnout. Survey scores for burnout, workload and exhaustion suggested improvement with both tools. The authors say these results need confirming in larger trials.
- Errors. Doctors said the scribes made clinically significant mistakes “occasionally”. One mild adverse event was reported.
A second randomised trial, at the University of Wisconsin, gave AI scribes to 66 practitioners in turn over 24 weeks (Afshar and colleagues, NEJM AI, 2025). Practitioners used the scribe for 38% of 71,487 notes. Time on notes fell by 0.36 hours a day. Work exhaustion fell. The fall in work done outside working hours did not hold once the most extreme 3% of days were left out.
When the result holds
The scribe was assigned to doctors, and they still chose visit by visit whether to use it. The trial did not record why.
The two tools gave different results in the same clinics. A result for one product says little about another.
Limits
- The trial lasted two months, in one health system.
- Burnout and accuracy were self-reported in surveys.
- The work is clinical documentation. Drafting notes from recorded meetings in other jobs is similar but untested.
- The trial was funded by the UCLA Department of Medicine and others, not by the vendors.
Practices that use this study
This study supports: