Background: Healthcare systems are under growing pressure to increase clinician productivity while reducing burnout. After-hours electronic health record (EHR) documentation- "work outside of work" (WOW)-is a major contributor to clinician fatigue. While generative AI-powered ambient clinical documentation (ACD) tools offer promise, their real-world impact remains uncertain.
Objective: To evaluate whether a hybrid ACD program (combining generative AI and a virtual scribe) is associated with changes in (1) WOW, (2) documentation delay, and (3) clinician financial productivity.
Design: Retrospective study (July 1-December 30, 2024).
Participants: In total, 181 attending physicians and advanced practice clinicians across 14 adult primary care practices within a Massachusetts healthcare system. Sixty-six clinicians (36%) voluntarily enrolled in the hybrid ACD program, which was rolled out using an activation schedule.
Intervention: The ACD tool used large language models to generate draft clinical notes that were then reviewed and finalized by a human scribe and integrated into the EHR. Participants received standardized training and technical support.
Main measures: Outcomes over the 50-day post-activation period included (1) WOW (5:30 pm-7:00 am weekdays; all weekends/holidays); (2) documentation delay (proportion of notes open ≥ 2 business days); and (3) financial productivity (work relative value units, wRVUs). Generalized estimating equations were used, adjusting for clinician and panel characteristics.
Key results: ACD adoption had a dynamic impact. WOW initially increased by 32.1% (95% CI 25.5%, 39.1%) on day 10 but decreased by 41.7% (95% CI -51.8%, -29.4%) by day 50 compared to baseline. Documentation delay consistently improved, with the odds of delayed note closure reduced by 66.5% (95% CI -68.3%, -64.7%) at day 10 and 66.0% (95% CI -72.5%, -57.8%) at day 50. Financial productivity increased steadily by 1.5% (95% CI -0.2%, 3.2%) on day 10 and 12.1% (95% CI 5.7%, 18.9%) by day 50.
Conclusion: ACD adoption was associated with reduced after-hours work, faster note completion, and improved financial productivity within 50 days of implementation.
Keywords: artificial intelligence; primary care; well-being.
© 2025. The Author(s), under exclusive licence to Society of General Internal Medicine.