AI scribe use associated with reduced documentation time but longer clinical notes

1. In this quality improvement study, ambient artificial intelligence (AI) scribes were associated with modestly reduced median time spent documenting clinician notes and increased median note length.

2. Clinicians who used AI scribes the most experienced the largest reduction in note-active time and the largest increase in note length compared with less frequent users and nonusers.

Evidence Rating Level: 2 (Good)

Study Rundown: Ambient AI-powered scribes have been marketed to reduce documentation burden and alleviate clinician burnout. These applications use speech input from patient-clinician encounters and large language models to generate clinical documentation. However, the evidence on AI scribes and clinician burnout remains mixed, with some studies showing positive effects and others reporting minimal impact on burnout as well as an increased need to correct AI-generated inaccuracies. This study aimed to assess the association between Heidi, an AI scribe, and clinical documentation outcomes. Heidi was more frequently used among clinicians in primary care and among clinicians with a higher monthly encounter volume and longer note-active time at baseline. Following Heidi enrollment, median note-active time modestly decreased and median note length increased, whereas same-day and after-hours note closure remained at similar levels. Adjusted interrupted time series (ITS) models showed generally consistent results. Clinicians who used Heidi the most had the largest immediate decrease in note-active time and a significant immediate increase in note length compared with low-frequency users and nonusers. Less frequent Heidi users experienced an immediate decrease in note length, while nonusers had no significant immediate change in note length. No significant between-group differences were observed in same-day or after-hours note closure. The study’s findings are limited by potential unmeasured confounding, an inability to distinguish between AI-generated and clinician-written notes, and a lack of assessment of note content and potential errors. Nevertheless, this study suggests that frequent AI scribe use may reduce note-active time, although its impact on note accuracy remains unclear.

Click to read this study in AIM

Relevant Reading: Clinical Implementation of Artificial Intelligence Scribes in Health Care: A Systematic Review

In-Depth [retrospective quality improvement study]: This quality improvement study aimed to assess the association between Heidi, an AI scribe, and clinical documentation burden. Physicians, nonphysician practitioners, trainees, and allied health professionals within a 14-hospital health system were eligible for this study. Data from encounters between July 1, 2024, and January 31, 2026, were collected. Only outpatient clinicians delivering direct patient care were included, and notes that were incomplete, unsigned, deleted, or associated with patients younger than 18 years were excluded. The primary outcome was note-active time per encounter, quantified as minutes spent with notes open in the electronic health record (EHR). A total of 1,140 clinicians enrolled in Heidi between January and November 2025. After excluding clinicians with fewer than 60 encounters prior to enrollment (n = 164), fewer than 3 months of Heidi enrollment (n = 55), or fewer than 30 encounters during their first 3 months of Heidi enrollment (n = 19), 902 clinicians met the study inclusion criteria. This cohort was divided into 302 (33.5%) high users, 303 (33.6%) low users, and 297 (32.9%) enrolled nonusers. Following Heidi enrollment, median note-active time decreased from 21.0 minutes (interquartile range [IQR], 8.9 to 38.4 minutes) to 18.6 minutes (IQR, 7.1 to 36.0 minutes), and median note length increased from 4,854 characters (IQR, 2,944 to 7,744 characters) to 5,187 characters (IQR, 3,144 to 8,188 characters). In adjusted ITS analyses, Heidi enrollment was associated with an immediate 1.21-minute reduction (95% CI, −1.74 to −0.68 minutes) in note-active time. No significant immediate change was observed in note length (8.11 characters [95% CI, −49.47 to 65.68 characters]), but note length increased significantly over time (25.08 characters per month [95% CI, 4.62 to 45.53 characters per month]). High-frequency Heidi users had the largest immediate reduction in note-active time (−2.50 minutes [95% CI, −3.55 to −1.45 minutes]). High users also experienced a significant immediate increase in note length (160.26 characters [95% CI, 42.70 to 277.82 characters]), whereas note length decreased significantly among low users (−110.44 characters [95% CI, −189.43 to −31.45 characters]) and did not significantly change among nonusers (−10.63 characters [95% CI, −105.67 to 84.41 characters]). No differences were observed in work relative value units (wRVUs). Overall, this study suggests that Heidi enrollment was associated with modest reductions in note-active time and increases in note length, with findings more pronounced among frequent users.

Citation: Booth SE, Gould R, Cibotti N, Fields R, Horng S, Jegadeesan V, et al. Impact of Unrestricted Artificial Intelligence Scribe Deployment in a Single Health System on Clinician Workload and Productivity : A Quality Improvement Study. Ann Intern Med. 2026 Oct 26. doi: 10.7326/annals-25-05294. Epub ahead of print. PMID: 42932880.

© 2026 2 Minute Medicine, Inc. All rights reserved. No part of this article may be reproduced, distributed or transmitted in any form or by any means without the prior written permission of 2 Minute Medicine, Inc. To license this content, contact 2 Minute Medicine, Inc. This article is for informational and educational purposes only; it is not medical advice and does not replace the judgment of a qualified healthcare professional.
Image: © 2 Minute Medicine