Ambient artificial intelligence scribes in the emergency department

In the February 2026 issue of Annals of Emergency Medicine, Preiksaitis et al. report the results of a single-center retrospective analysis of voluntary digital scribe (DAX Copilot) adoption in a tertiary emergency department (ED) (1). They found that digital-scribe-assisted encounters were associated with lower on-shift documentation time {165 [97–250] seconds vs. 230 [142–365] seconds} and total electronic health record (EHR) time {519 [349–751] seconds vs. 621 [412–914] seconds}. Their findings extend a growing ambient-scribe literature that remains driven largely by ambulatory and outpatient studies, with comparatively limited evidence from the ED (2-10). Among ED-specific digital-scribe studies, only the recent multicenter analysis by Dutta et al. is meaningfully larger than the analysis by Preiksaitis et al (1,7-10). In this limited evidence base, Preiksaitis et al. should be interpreted as a real-world study of voluntary digital-scribe adoption in a distinctive implementation setting, rather than as causal evidence that digital scribes reduce ED documentation time.

Preiksaitis et al. diligently acknowledge several important caveats to their work and frame it as an early implementation report (1). Ambient scribe use was limited to 976 of 8,740 eligible encounters (11.2%) during the 8-month study period, and adoption was highly concentrated: 9 of 92 physicians (9.8%) accounted for 70.5% of all scribe-assisted encounters. As a result, the study provides somewhat less robust evidence than the multicenter analysis by Dutta et al., which included more than 8,000 digital-scribe encounters across 38 physicians (7). Nevertheless, although no randomized controlled trials have evaluated ambient scribe use specifically in the ED setting, the convergence of several favorable observational studies cautiously suggests that digital scribes may have utility in emergency care (1,7-10). Interpreting that literature requires careful attention to the comparison group. Most ED studies have compared digital scribes with conventional documentation or compared one ambient platform with another (1,7,9,10). The only ED study that directly compared digital scribes with human scribes found superior performance with human scribes (8). Because human scribes are generally more expensive than digital scribes, future trials should evaluate digital scribes against both conventional documentation and human scribes, incorporating documentation time, note quality, clinician wellbeing, and cost (11-14). Evaluating clinician wellbeing in such a study is particularly important because even modest reductions in documentation time (1.5 hours per week) have been associated with large reductions in clinician burnout in some ambulatory settings (3). A one-minute reduction in documentation time per encounter is less operationally compelling than evidence that digital scribes improve the working lives of frontline clinicians.

Given the limited number of participants, one might frame the results of Preiksaitis et al. as the initial report of a highly selected group of early digital scribe innovators in a unique setting. This is particularly important because digital scribe adoption is not evenly distributed across documentation styles. At our own institution, early digital scribe adopters tend to be more willing to accept variation in narrative than providers who prefer conventional documentation. The latter often tend to write lengthier and more time-consuming notes. A more robust study design could have leveraged both ambient and standard encounters during the same implementation period and constructed a mixed-effects model, adjusting for encounter-level covariates such as Emergency Severity Index (ESI), care zone, disposition, interpreter need, and physician-level random effects. Such a design would have reduced confounding by case selection and helped mollify impacts of clinician-level variation in note style. A complementary interrupted time-series analysis of shift-level documentation time before and after scribe uptake could also have addressed a more pragmatic implementation question: whether making digital scribes available for selected encounters reduces overall clinician documentation burden across a heterogeneous ED shift (3).

In the Preiksaitis et al. study, digital scribe use was also heavily concentrated in low-acuity and telemedicine encounters. Though Preiksaitis et al. attempted to ameliorate these differences by stratifying encounters based on ESI, such stratification almost certainly leaves substantial residual confounding between the “standard” and “ambient artificial intelligence (AI)” encounter groups. It is difficult to imagine that providers would require the same amount of documentation time for a patient treated in an urgent-care-style vertical care setting and discharged home as for a patient presenting to traditional emergency care and requiring inpatient admission, even if both encounters had an initial ESI acuity level of 3. Thus, whether the time savings seen in Preiksaitis et al.’s work are truly associated with digital scribe use or visit characteristics remains uncertain. Given the real-world nature of the work, perfectly accounting for such encounter-level selection is impossible, though the methods above may have offered more insight.

A related constraint is that the study included only single physician encounters. The astute reader is likely to wonder how this criterion might bias case mix. In a tertiary academic ED, restricting analyses to single-attending encounters may exclude a meaningful subset of higher-complexity encounters, including long-stay visits involving attending handoffs, trainees, or resuscitation workflows. The magnitude and direction of this bias will vary by institutional staffing and supervision model, but the single-attending criterion likely limits generalizability to the interrupted, handoff-heavy, high-acuity encounters that distinguish ED practice. It may also partially explain why there were fewer ESI 2 visits included in the Ambient AI group (6.8% compared to 14.8% of standard encounters). Whether digital scribes are of benefit to trainees remains an important question for future work. Taken together, the high proportion of low-acuity vertical care visits and the exclusion of trainees likely makes the Preiksaitis study more representative of a trial in a “fast track” or “urgent care” environment (1). To that end, other studies have already noted that ambient scribes are particularly useful in outpatient urgent care visits (15). While the work of Preiksaitis et al. represents a reasonable pilot for low-acuity visits, more work is needed to understand the benefit of digital scribes in the highest acuity and complexity encounters, which distinguish Emergency Medicine from other acute care settings. The potential for ambient documentation in the emergency setting will not be realized until the full breadth of patient encounters can be captured.

Like some of the prior ambient-scribe literature, Preiksaitis et al. relied predominantly on EHR audit-log measures (1,7). Many academic informatics projects rely on readily obtainable EHR data. These data offer the benefit of automated objective collection with nearly limitless scale, standardized across organizations for easy comparison. However, they are notably limited in their ability to capture clinician documentation burden. For this reason, many studies of ambient scribes are now combining EHR records with self-reported total documentation time, after-shift or remote work, cognitive load, and clinician experience (2-5,8-10). First, many ambient scribes (including DAX) allow clinicians to edit their notes in a web-based portal that may not be captured in EHR activity records (16). Second, some commercial platforms add time before and after a scheduled appointment that is considered “on shift” (17-19). Third, these platforms often use a 5-second inactivity window to determine whether a clinician is actively working on documentation. This begs the question: if a clinician spends 15 seconds contemplating how best to document an encounter, should this not be included as documentation time? Whether documentation is a cognitive task or merely a series of keystrokes thus seems to remain unresolved between clinicians and EHR developers. However, the substantial reductions in cognitive burden reported in other studies of ambient scribes suggest that clinicians likely experience significant intellectual effort that is alleviated by these tools (2-6). Similarly, discrepancies in clinician-reported time savings or EHR metrics across multiple prior studies should prompt us to ask: who is the best adjudicator of clinician time? (3,6,20,21) These important questions remain unanswered, and should prompt further discussion in the field.

Statistical design aside, the very low adoption rate in Preiksaitis et al.’s study highlights an important challenge in clinical informatics: scaling transformative technologies. In ambulatory settings, most evidence suggests that ambient scribes reduce burnout and cognitive burden and improve work-life integration and patient engagement (3-5,22-24). Two ED-specific studies now suggest similar perceived benefits for some emergency physicians (9,10). In these small studies, most ED physicians reported being satisfied with the ambient scribe and reductions in perceived work burden (9,10). However, ambient scribes are imperfect; ambulatory providers experience barriers including device availability, language access, and editing time (24). Similarly, ED providers identify additional setting-specific barriers including environmental noise, complex patient populations, workflow disruption, limited trust in AI-generated notes, and limited utility for physical-examination and medical-decision-making documentation (9,10). New features such as context-aware decision support and custom prompt engineering may serve as important accelerators of clinician adoption of digital scribes (25). However, informaticians and clinical leaders must acknowledge that most new technologies (even those specifically designed for clinical efficiency) are transiently associated with decreased productivity before long-term benefits can be realized. Dedicated time for clinical implementation must be considered for any rollout to be successful. Ambient documentation is a learned skill, much like conventional documentation. Lessons from ambulatory digital scribe implementations will be useful, but they translate imperfectly to the ED, where documentation is often nonlinear and medical decision-making evolves as new information emerges during the encounter. Such specialty-specific details often require manual edits in ambient documentation (26). Further, ED documentation is complicated by background noise, frequent interruptions, and incomplete patient participation—particularly in high-acuity encounters—all of which make passive capture more difficult. Successful ED implementation will therefore require dedicated time for skill acquisition, workflow adaptation, and model customization that reflects the distinctive environment of emergency care. Without this support, future implementations may see only a handful of providers embrace digital scribes, as was the case in Preiksaitis et al (1).

Overall, we applaud the important work of Preiksaitis et al. in evaluating ambient-scribe use in the dynamic and clinically distinct environment of a tertiary academic ED. Their study adds to the small but growing body of ED-specific evidence, while also contributing to a broader literature suggesting that ambient digital scribes may reduce documentation time, decrease cognitive burden, and improve clinician wellbeing across multiple care settings. Future ED ambient-scribe studies should move beyond pilot implementation reports based primarily on EHR-derived time metrics. Randomized trials are needed, ideally comparing ambient scribes with both conventional documentation and human scribes. These trials should be designed to evaluate performance not only in shorter, lower-acuity encounters, but also in high-acuity, interrupted, and handoff-heavy encounters that define emergency care. Key outcomes should include documentation time, after-shift work, note quality, clinician burden, patient experience, and cost.

Acknowledgments

Artificial intelligence was used for literature searches and spelling/grammar checking.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Journal of Medical Artificial Intelligence. The article has undergone external peer review.

Peer Review File: Available at https://jmai.amegroups.com/article/view/10.21037/jmai-2026-0082/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jmai.amegroups.com/article/view/10.21037/jmai-2026-0082/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Preiksaitis C, Alvarez A, Winkel M, et al. Ambient Artificial Intelligence Scribe Adoption and Documentation Time in the Emergency Department. Ann Emerg Med 2026;87:569-74. [Crossref] [PubMed]
2. Olson KD, Meeker D, Troup M, et al. Use of Ambient AI Scribes to Reduce Administrative Burden and Professional Burnout. JAMA Netw Open 2025;8:e2534976. [Crossref] [PubMed]
3. Pelletier JH, Watson K, Michel J, et al. Effect of a generative artificial intelligence digital scribe on pediatric provider documentation time, cognitive burden, and burnout. JAMIA Open 2025;8:ooaf068. [Crossref] [PubMed]
4. Lukac PJ, Turner W, Vangala S, et al. Ambient AI Scribes in Clinical Practice: A Randomized Trial. NEJM AI 2025;2: [Crossref] [PubMed]
5. Stults CD, Deng S, Martinez MC, et al. Evaluation of an Ambient Artificial Intelligence Documentation Platform for Clinicians. JAMA Netw Open 2025;8:e258614. [Crossref] [PubMed]
6. Hassan H, Zipursky AR, Rabbani N, et al. Corrigendum: Clinical Implementation of Artificial Intelligence Scribes in Health Care: A Systematic Review. Appl Clin Inform 2025;16:e1. [Crossref] [PubMed]
7. Dutta S, Guan-Ting You J, Dunham L, et al. Medical Scribe and Ambient Artificial Intelligence Impact on Emergency Physician Documentation Burden and Clinical Productivity. Ann Emerg Med 2026; Epub ahead of print. [Crossref]
8. Morey J, Jones D, Walker L, et al. Ambient Artificial Intelligence Versus Human Scribes in the Emergency Department. Ann Emerg Med 2026;87:561-8. [Crossref] [PubMed]
9. Marquis T, Kopp M, Anderson JS, et al. AI-Powered Ambient Scribe Technology Experiences Among Emergency Physicians: Cross-Sectional, Mixed Methods Pilot Survey Study. JMIR Form Res 2026;10:e80401. [Crossref] [PubMed]
10. Webb J, Chu L, Turer RW, et al. Crossover Evaluation of Two Ambient AI Scribe Tools in the Emergency Department. Appl Clin Inform 2026;17:118-26. [Crossref] [PubMed]
11. Heaton HA, Nestler DM, Barry WJ, et al. A Time-Driven Activity-Based Costing Analysis of Emergency Department Scribes. Mayo Clin Proc Innov Qual Outcomes 2019;3:30-4. [Crossref] [PubMed]
12. Nong P, Neprash HT. Unintended Consequences of Using Ambient Artificial Intelligence Scribes for Billing. JAMA Health Forum 2026;7:e255771. [Crossref] [PubMed]
13. Shah SJ, Garcia P, Ambient AI. Scribes-What Is the Return on Investment? JAMA Netw Open 2026;9:e2553238. [Crossref] [PubMed]
14. Miksanek TJ, Skandari MR, Ham SA, et al. The Productivity Requirements of Implementing a Medical Scribe Program. Ann Intern Med 2021;174:1-7. [Crossref] [PubMed]
15. Rabbani N, Ray M, Verhagen E, et al. Ambient Artificial Intelligence Scribes in Pediatric Primary Care: A Mixed Methods Study. Appl Clin Inform 2025;16:1578-87. [Crossref] [PubMed]
16. Microsoft. DAX copilot for epic FAQs. n.d. Available online: https://support.microsoft.com/it-it/dax-copilot/epic/dax-copilot-for-epic-faqs
17. Sinsky CA, Trockel MT, Dyrbye LN, et al. Vacation Days Taken, Work During Vacation, and Burnout Among US Physicians. JAMA Netw Open 2024;7:e2351635. [Crossref] [PubMed]
18. Arndt BG, Beasley JW, Watkinson MD, et al. Tethered to the EHR: Primary Care Physician Workload Assessment Using EHR Event Log Data and Time-Motion Observations. Ann Fam Med 2017;15:419-26. [Crossref] [PubMed]
19. Rule A, Melnick ER, Apathy NC. Using event logs to observe interactions with electronic health records: an updated scoping review shows increasing use of vendor-derived measures. J Am Med Inform Assoc 2022;30:144-54. [Crossref] [PubMed]
20. Tajirian T, Stergiopoulos V, Strudwick G, et al. The Influence of Electronic Health Record Use on Physician Burnout: Cross-Sectional Survey. J Med Internet Res 2020;22:e19274. [Crossref] [PubMed]
21. Gilleland M, Komis K, Chawla S, et al. Resident duty hours in the outpatient electronic health record era: inaccuracies and implications. J Grad Med Educ 2014;6:151-4. [Crossref] [PubMed]
22. Duggan MJ, Gervase J, Schoenbaum A, et al. Clinician Experiences With Ambient Scribe Technology to Assist With Documentation Burden and Efficiency. JAMA Netw Open 2025;8:e2460637. [Crossref] [PubMed]
23. You JG, Dbouk RH, Landman A, et al. Ambient Documentation Technology in Clinician Experience of Documentation Burden and Burnout. JAMA Netw Open 2025;8:e2528056. [Crossref] [PubMed]
24. Shah SJ, Crowell T, Jeong Y, et al. Physician Perspectives on Ambient AI Scribes. JAMA Netw Open 2025;8:e251904. [Crossref] [PubMed]
25. Troup M, Asawa C. From ambient AI to clinician-directed intelligence: How prompt editing and contextual decision support will redefine documentation. 2026. [cited 2026 Apr 10]. Available online: https://www.abridge.com/blog/ambient-ai-prompt-editing
26. Guo Y, Hu D, Yang Z, et al. What do clinicians edit in ambient AI-drafted clinical documentation? A qualitative content analysis. J Am Med Inform Assoc 2026;ocag073.