A UCSF-led study published in JAMA Network Open has quantified what AI scribe vendors have long claimed anecdotally: physicians who adopt ambient AI documentation tools see measurably higher productivity and revenue than those who don’t, though the gains are modest enough that researchers are urging caution before treating them as a guaranteed return on investment.
A Large, Real-World Sample
The research team, led by A Jay Holmgren of UCSF’s Division of Clinical Informatics and Digital Transformation, analyzed nearly 1.2 million ambulatory encounters across 1,565 physicians between January 2023 and April 2025. Of those physicians, 698 — about 44.6% — adopted AI scribe technology during the study period, while 867 remained non-adopters, giving the researchers a large natural comparison group rather than a small controlled trial.
The Numbers Behind the Headline
Physicians who adopted AI scribes generated 1.81 more relative value units (RVUs) per week than non-adopters, a 5.8% increase that the authors estimate translates to roughly $3,044 in additional revenue per physician per year, using 2025 Medicare billing rates. Adopters also handled 0.80 more patient encounters per week, a 2.8% rise in visit volume, suggesting at least part of the productivity gain came from physicians seeing more patients rather than purely from improved documentation efficiency.
No Sign of a Billing Red Flag
One of the more reassuring findings for hospital administrators: there was no difference in the proportion of insurance claims denied between AI-scribe adopters and non-adopters. That matters because critics of ambient documentation tools have warned that AI-generated notes could trigger more coding errors or claim pushback, and this dataset found no such effect over the study period.
What the Study Couldn’t Pin Down
The authors acknowledge they could not determine whether the higher RVUs came from physicians delivering more clinical services, or simply documenting and coding existing visits more accurately and completely than before. That ambiguity matters for how hospital systems should interpret the return: if the gain is mostly better coding capture, the revenue bump may not reflect any change in actual patient care delivered.
A Modest Gain, by the Researchers’ Own Framing
An accompanying editorial from Stanford authors described the per-physician financial gains as “modest,” a characterization that cuts against some of the more aggressive ROI claims made by AI scribe vendors marketing the tools to health systems. Separately, a related UCSF and Mass General Brigham analysis found AI scribe users spend about 30 fewer minutes per day inside the EHR and add roughly one additional visit every two weeks — a smaller but directionally consistent signal.
Balancing Enthusiasm and Caution
Hospital technology leaders have pointed to this study as validation for scribe rollouts that were previously justified mostly by physician satisfaction surveys rather than hard revenue data. Skeptics counter that a 5.8% RVU bump and roughly $3,000 a year per physician is a thin margin against the subscription costs of enterprise AI scribe platforms, and that health systems should demand clearer evidence on whether the gains come from genuine productivity or coding optimization before expanding contracts.
Looking Ahead
With adoption already above 60% at many Epic-using hospitals nationally, the open question is whether gains like these compound as physicians get more comfortable with the tools, or plateau once the easiest documentation efficiencies have been captured. Future studies tracking the same physician cohorts over a longer period should clarify which scenario is playing out.
Photo: Pexels / PIXABAY via Pixabay
