Insights
Do AI scribes actually give practitioners time back?
Yes — but less than the sales deck promises, and not in the place you would expect. The largest study to date, covering roughly 1,800 clinicians across five academic medical centers, found about 16 minutes of documentation time saved for every eight hours of patient care. Real, and modest. The gain that shows up largest and most consistently is not minutes on a clock. It is what happens to a practitioner's attention when they are no longer typing while someone tells them the hardest thing they have said all week.
How much time do AI scribes actually save?
About 16 minutes of documentation time per eight hours of patient care, plus 13 fewer minutes inside the medical record. That is what the largest study to date reported — published in JAMA, following roughly 1,800 clinicians at five academic medical centers from 2023 to 2025. It also found something worth sitting with: no significant change in time spent in the record outside working hours. The evenings did not come back on their own. Scribe users could see about one additional patient every two weeks — a genuine number, and a smaller one than most practices are told to expect. Primary care practitioners benefited more than others.
From the field
A large group practice we work with ran ambient documentation for a full quarter before deciding anything. The stopwatch numbers came in close to what the research predicts — underwhelming, if minutes were the point. What changed the conversation was a different question: what happens in the room now? Practitioners described finishing visits facing the person instead of the screen. Two of them still turned the tool off, and they had good reasons — one worked where the ambient audio was unreliable, and one found that correcting the notes cost more than writing them. The practitioners who kept it were, almost without exception, the ones who had been coached through the first two weeks rather than handed a login.
If the minutes are modest, why do practitioners say it changed their day?
Because the burden a scribe lifts is cognitive before it is chronological. In a 2025 JAMA Network Open study of 263 physicians and nonphysician providers across six health systems, burnout among those in ambulatory clinics fell from 51.9% to 38.8% after 30 days with an ambient scribe, alongside improvements in cognitive task load and in focused attention on the patient. A separate JAMA Network Open study published August 21, 2025, surveying more than 1,400 physicians and advanced practice providers, found a 21.2% absolute reduction in burnout prevalence at Mass General Brigham by day 84. The qualitative feedback is the part we keep returning to: practitioners described more contact with patients and families, and a return of joy in practice. Presence is not a soft benefit. A practitioner who is fully in the room hears the thing that was not on the intake form — and people who feel heard stay, refer, and follow through.
51.9% → 38.8%
Burnout among ambulatory practitioners after 30 days with an ambient scribe (JAMA Network Open, 2025). The honest limits: 263 participants, self-reported surveys, no randomized control group, and modest response rates in the related multi-site work. The direction is consistent across studies. The size of the effect is not.
What separates a scribe that helps from one that gets switched off?
Two habits, neither of them technical. The first is coaching. The same large study that found modest time savings also found inconsistent use, and concluded that clinicians need guidance to get the most from these tools — the benefit tracks with how a tool is taught, not merely whether it is installed. The second is a review habit. A 2026 narrative review of ambient scribe research reports frequent documentation omissions and intermittent factual inaccuracies, with physical examination documentation flagged as the highest-risk area — systems have generated exams that never happened. The note carries your signature and your judgment. Read it the way you would read a note written by a colleague who was not in the room, because that is exactly what it is.
What should be settled before the first recording?
The vendor agreement. If a tool will hear, process, or store patient information on your behalf, it is a business associate under HIPAA, and a signed business associate agreement belongs in place before the first visit is recorded — not after the pilot goes well. Many general-purpose AI assistants will not sign one at all, and that is your answer about whether they belong anywhere near a treatment room. Settle two more things while you are there: what the vendor may do with your recordings and notes beyond serving you, and what you will measure at 30, 60, and 90 days. Choose the measure before you start. The AMA's 2025 survey of nearly 19,000 physicians is a useful reminder of why: burnout has fallen three years running, to 41.9%, while the share of physicians spending more than eight hours a week in the record after hours has not moved since 2022. Tools alone do not move that number. Workflow does.
The honest case for an ambient scribe was never the minutes. It is that a practitioner who is not typing can look up. That is where we start with every system we build — the technology is only worth what it gives back to the space between a practitioner and the person in front of them. If you are evaluating one, measure the room, not only the clock.