AI Scribes Reduce Documentation Burden but Deliver Modest Gains in Efficiency, Multisite Study Finds

A large, multisite study evaluated whether AI-powered medical scribes meaningfully reduce clinician workload and improve productivity. The researchers examined real-world adoption of ambient AI documentation tools across five major U.S. academic health systems. Overall, the findings published in JAMA suggest that while AI scribes do reduce time spent on electronic health records (EHRs), the benefits are modest and unevenly distributed across clinician groups.

Study Details

This longitudinal cohort study analyzed 8,581 ambulatory clinicians, including 1,809 who adopted AI scribes, across five academic institutions that implemented these tools between June 2023 and August 2025.  Across the participating institutions, clinicians mostly opted-in to AI adoption once they were deemed eligible. 

The researchers used a difference-in-differences design to compare clinicians’ performance before and after adoption with that of clinicians who did not adopt the technology.

AI scribe adoption was defined as receiving access to tools such as Abridge, Nuance DAX Copilot, or Ambience, all integrated with Epic EHR systems. The study focused on several key outcomes: total time spent in the EHR, time devoted specifically to documentation, time spent working in the EHR outside scheduled hours, and weekly patient visit volume. All time-based measures were normalized to an 8-hour clinical day to allow for consistent comparisons.

Key Takeaways

The study found that AI scribe adoption was associated with statistically significant but modest improvements in clinician efficiency. On average, clinicians who adopted AI scribes spent 13.4 fewer minutes (95% confidence interval [CI] = 9.1–17.7 minutes) in the EHR and 16.0 fewer minutes (95% CI = 13.7–18.3 minutes) on documentation. At the same time, they completed approximately 0.49 additional patient visits per week. There was no statistically significant reduction noted in time spent working in the EHR outside of work hours.

The magnitude of benefit varied across clinician groups. Primary care clinicians, advanced practice clinicians, female clinicians, and those who used AI scribes in at least half of their patient encounters experienced the greatest improvements. In particular, high-frequency users of AI scribes saw substantially larger reductions in documentation time and more pronounced increases in visit volume compared with lower-use clinicians.

From a financial perspective, the study estimated that AI scribe adoption was associated with a modest increase in revenue, averaging about $167 per clinician per month. This increase likely reflects a combination of slightly higher visit volume and improved documentation supporting billing.

Study authors noted that as a result of the nonrandomized nature of the study, associations could have been due to AI scribe implementation, or other factors that were not measured in the study.  

The authors, including lead study author Lisa S. Rotenstein, MD, MBA, MSc, of the Division of Clinical Informatics and Digital Transformation at the University of California, San Francisco, summarized their overall findings by noting that “AI scribe adoption was associated with modest decreases in total EHR time and documentation time and with a modest increase in weekly visit volume.” They concluded that future studies of the impact of AI scribes would need to address specific workflows to determine how AI could further improve clinician efficiency. 

Another Viewpoint

In a related editorial published in the same issue of JAMA, Aaron A. Tierney, PhD, and colleagues from Kaiser Permanente, placed these results in a broader context, emphasizing that most current evaluations of AI scribes focus on easily measurable metrics, such as time savings and productivity. The editorial argued that the more important question is how the saved time is used, stating that “the question is no longer whether ambient AI can save documentation time…but whether that time is reinvested in ways that measurably improve outcomes and equity for patients.”

The editorial also highlighted key gaps in the evidence, including limited data on patient experience, population health outcomes, and equity. Although clinician workflow improvements are increasingly well documented, the downstream effects on care quality and health disparities remain uncertain.

Conclusion

This multisite study provides one of the most comprehensive real-world evaluations of AI scribes to date. It confirms that these tools can reduce documentation burden and modestly increase productivity, but the overall magnitude of benefit is limited and varies across users. The findings suggest that consistent use and targeted deployment may be necessary to realize the greatest gains.

At the same time, efficiency metrics alone are insufficient to assess the value of AI in health care. As AI scribes become more deeply embedded in clinical practice, future research will need to focus on whether these tools meaningfully improve patient outcomes, clinician well-being, and health equity, rather than simply reducing time spent on documentation.

DISCLOSURES: Dr. Rotenstein, Dr. Holmgren, Mr. Thombley, and Ms. Sriram were supported by a grant from the Advancing a Healthier Wisconsin Endowment. Dr. Adler-Milstein reported that this work was supported in part by a gift from Ken and Kathy Hao to establish the Impact Monitoring Platform for AI in Clinical Care at the University of California, San Francisco. Ms. Iannaccone and Ms. Frits were supported by an Agency for Healthcare Research and Quality (AHRQ) grant. For full disclosures of the study authors, visit jamanetwork.com.

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.