AI Scribes Are Starting to Ease Physician Burnout and Cut Documentation Time According to a New UCLA Trial
Artificial intelligence tools designed to automatically document doctor–patient visits are beginning to show real, measurable benefits in clinical practice. A new randomized clinical trial conducted at UCLA Health suggests that AI-powered medical scribes can reduce documentation time, improve aspects of physician well-being, and potentially help address one of modern medicine’s most persistent problems: documentation overload.
The study, published in NEJM AI, is one of the first rigorous, real-world trials to evaluate whether AI scribes actually deliver on their promise outside of controlled demos and pilot programs. With health systems across the U.S. rapidly adopting these tools, the findings arrive at a particularly important moment.
Why Documentation Has Become Such a Big Problem for Doctors
Over the past decade, electronic health records (EHRs) have transformed how medical care is delivered and documented. While EHRs have improved access to information and coordination of care, they have also dramatically increased the amount of time physicians spend on administrative work.
Multiple studies have shown that physicians often spend nearly two hours on documentation for every hour of direct patient care. This imbalance has become a major contributor to physician burnout, work dissatisfaction, and even early retirement. Burnout now affects close to half of U.S. physicians, contributing to workforce shortages, increased medical errors, and billions of dollars in costs for health systems.
AI scribes are designed to address this issue by listening to clinical conversations, generating draft notes automatically, and allowing physicians to review and finalize them. The UCLA trial set out to determine whether this approach truly makes a difference when used at scale.
How the UCLA AI Scribe Study Was Designed
The trial was conducted at UCLA Health between November 2024 and January 2025. Researchers enrolled 238 physicians across 14 different medical specialties, covering approximately 72,000 patient encounters. Participants were randomly assigned to one of three groups:
- Physicians using Nabla, an AI scribe application
- Physicians using Microsoft Dragon Ambient eXperience (DAX)
- A control group continuing with usual documentation practices
This randomized design is significant. Many AI tools are adopted based on anecdotal enthusiasm or vendor claims, but randomized trials remain the gold standard for determining whether a technology truly works.
Both AI scribes operated in a similar way: they recorded patient–physician conversations during visits and generated draft clinical notes, which physicians then reviewed and edited before finalizing.
What the Study Found About Documentation Time
One of the most closely watched outcomes was how much time physicians spent writing notes.
Physicians using Nabla experienced a statistically significant reduction in documentation time. On average, time spent writing each note dropped by 41 seconds, decreasing from 4 minutes and 30 seconds to 3 minutes and 49 seconds. Compared with the control group, this represented a 9.5% larger reduction, which reached statistical significance.
In contrast, physicians using Microsoft DAX saw a smaller decrease of 18 seconds per note, dropping from 4 minutes and 22 seconds to 4 minutes and 4 seconds. While this change moved in a positive direction, it did not reach statistical significance compared to the control group.
These time savings may seem modest at first glance, but when multiplied across dozens of patient visits per day and hundreds per month, they can add up to meaningful reductions in administrative burden.
Effects on Burnout, Workload, and Stress
Beyond time savings, the study examined how AI scribes affected physician well-being, using validated survey instruments to measure burnout, cognitive workload, and work exhaustion.
Physicians in both AI scribe groups reported modest but consistent improvements in burnout-related measures compared with the control group. On average, burnout scores improved by approximately 7% among physicians using Nabla or DAX.
Researchers also observed improvements in cognitive workload and work exhaustion, suggesting that even small reductions in documentation friction may have psychological benefits. Importantly, these improvements were seen even when time savings were relatively small, highlighting that how work feels can matter as much as how long it takes.
Accuracy Issues and Patient Safety Concerns
The study also identified important limitations that should temper enthusiasm.
Physicians reported that AI-generated notes occasionally contained clinically significant inaccuracies. The most common issues were omissions of relevant information and pronoun errors, which could potentially lead to confusion or misinterpretation if not corrected.
During the trial, one mild patient safety event was reported that was linked to AI-generated documentation. While no serious harm occurred, this finding reinforces a key message from the researchers: AI scribes require active physician oversight, not passive acceptance.
The study authors emphasized that AI-generated notes should always be treated as drafts, with clinicians maintaining full responsibility for accuracy and completeness.
Physician and Patient Reactions to AI Scribes
Survey responses revealed that most physicians found both AI scribe tools easy to learn and use. Many reported that the tools helped them stay more engaged with patients by reducing the need to type or focus on screens during visits.
Patients were also generally receptive. Fewer than 10% of patients declined the use of AI scribes during their visits, suggesting broad acceptance when the technology is clearly explained and used transparently.
Why This Study Matters for Health Care Systems
AI scribes are being adopted rapidly across hospitals, clinics, and private practices. However, many implementations occur without rigorous evaluation of effectiveness or safety. This UCLA study stands out because it embedded a randomized trial directly into routine clinical practice, generating real-world evidence rather than theoretical claims.
The researchers argue that this approach should become a model for evaluating other AI tools in health care, especially as generative AI becomes more deeply integrated into clinical workflows.
Limitations and the Need for Further Research
While the results are encouraging, the study has clear limitations. It was conducted at a single academic medical center, over a relatively short time period, and focused primarily on documentation-related outcomes.
The authors call for longer-term studies across multiple institutions to better understand how AI scribes affect downstream outcomes, including quality of care, health care costs, and patient experience.
What AI Scribes Could Mean for the Future of Medicine
AI scribes are unlikely to solve physician burnout on their own. Burnout is driven by a complex mix of workload, system design, staffing shortages, and organizational culture. However, this study suggests that AI scribes can be a meaningful part of the solution, especially when implemented thoughtfully and evaluated rigorously.
As AI tools continue to evolve, the challenge for health care systems will be balancing innovation with safety, and efficiency with clinical responsibility. This trial provides an important data point in that ongoing conversation.
Research Paper:
https://www.nejm.org/doi/10.1056/aioa2501000
