Forty-one seconds.
That was the average reduction in time spent writing each note for physicians using one of two ambient AI scribes in a randomised trial at UCLA Health, released as a preprint in July 2025. The control group, doing nothing new, shaved off 18 seconds anyway. The other scribe product in the trial showed a smaller drop that did not reach statistical significance.
Set that beside the other headline of the year. In October, JAMA Network Open published a study of 263 clinicians across six US health systems who started using an ambient scribe. After 30 days, the share reporting burnout fell from 51.9% to 38.8%.
A modest time saving, and a large drop in burnout. Both can be true. The gap between them is the most interesting thing about ambient AI, and it points to where the real risk lies.
The relief is real, and it is not mostly about minutes
Anyone who has spent an evening finishing notes knows the documentation burden is not imaginary. A time-and-motion study by Christine Sinsky and colleagues in the Annals of Internal Medicine in 2016 found that for every hour physicians spent face to face with patients, they spent nearly two more on EHR and desk work during the clinic day, plus one to two hours each night.
Ambient scribes attack the most visible part of that. At Kaiser Permanente, 7,260 physicians used the tool in more than 2.5 million encounters over roughly 14 months, according to a report in NEJM Catalyst. The organisation estimated more than 15,700 hours of documentation time saved over a year.
Spread across thousands of doctors, that works out to roughly two hours per physician over the year. Useful, not transformative.
So why do burnout scores move so much? The more plausible explanation is cognitive, not clock-based. Clinicians describe being able to look at the patient instead of the screen, and not carrying half-finished notes in their heads through the afternoon. The JAMA Network Open study reported reductions in cognitive task load as well as documentation time. Relief from divided attention may matter more than the minutes themselves.
That distinction matters, because the minutes are what administrators can see.
The uncomfortable incentive: saved time gets spent
Here is the trade-off that rarely makes the vendor slide deck. When a tool saves clinician time, the organisation paying for it has a strong incentive to convert that time into something billable. An extra appointment slot or two per session. A slightly shorter visit length. A larger panel.
This is not cynicism. It is how health system finance works. Ambient scribes are not cheap, and executives will be asked what return they produced. “Our doctors feel better” is a harder line item to defend than “we added capacity.”
If the time is recaptured as throughput, the burnout benefit is likely to shrink. The clinician is back to the same total load, with more patients, more orders and more follow-up messages generated by each additional visit. The note got easier. The day did not.
There is a second-order effect too. Scribes make it easier to generate longer, more complete notes. Longer notes can support higher billing codes. Payers are already watching coding intensity closely, and a documentation arms race tends to end with new audit rules and more administrative work, not less.
What the evidence does not yet show
The studies to date are encouraging but early. Most are pre-post comparisons without a control group, which means some of the improvement could reflect enthusiasm for anything new, the clinicians who volunteered, or regression from a particularly bad baseline. The UCLA trial is one of the first randomised comparisons, and its authors called for larger multicentre trials to confirm the well-being findings.
Other questions are still open:
- Accuracy. Clinicians still have to review every note. Errors of omission, or confidently stated details that were never said, carry liability that falls on the signer.
- Durability. Thirty days of relief is not twelve months of relief.
- Who benefits. Specialties with long, conversational visits may gain more than procedural ones. Clinicians working in a second language, or with patients who speak another, may see different accuracy.
- Consent and trust. Patients are being recorded. How that is explained, and whether some decline, will vary by setting.
Making the benefit stick
If health systems want scribes to reduce burnout rather than just speed up documentation, a few choices matter more than which product they buy:
- Protect the time explicitly. Decide in advance, and publicly, that saved time will not automatically become extra slots for a defined period.
- Measure burnout, not only minutes. Track well-being at six and twelve months, and share it with clinicians.
- Fix the inbox too. Documentation is one part of the EHR burden. Message volume, prior authorisation and results management are others, and a scribe does nothing for them.
- Watch the coding. Monitor whether billing levels shift after adoption, before a payer does it for you.
The mechanism: why burnout is about attention, not minutes
The gap between modest time savings and large drops in burnout makes more sense once you look at what burnout actually measures.
Burnout is usually described as a combination of emotional exhaustion, depersonalisation and a reduced sense of accomplishment. None of these is a simple function of hours worked. Clinicians who work long hours on meaningful tasks, with control over their day, often report less burnout than those who work fewer hours on tasks they see as pointless or imposed.
Documentation sits squarely in the second category for many clinicians. It is experienced as work done for billing, compliance and defence, not for the patient. It also competes directly with the patient for attention. Typing during a consultation means breaking eye contact, missing cues and feeling like a clerk rather than a clinician. An unfinished note hangs over the rest of the day.
An ambient scribe changes the texture of the encounter more than its length. The clinician can face the patient. The note begins to exist without effort. What remains is editing, which is cognitively lighter than composing. That may be why clinicians report relief out of proportion to the minutes saved.
This has an important implication. If the benefit comes from restoring attention and reducing the sense of doing clerical work, it can be undone by anything that reintroduces those burdens, including extra patients per session, more inbox work or new documentation requirements attached to the scribe itself.
A worked scenario: where the saved time goes
Take a primary care physician who sees 20 patients a day. Suppose an ambient scribe saves two minutes per note, which is more generous than the trial evidence so far. That is 40 minutes a day.
In the first version of the story, the organisation leaves the schedule alone. The physician finishes notes before leaving, stops logging in after dinner and has a few minutes between patients to think. Burnout scores fall and stay down. Retention improves. The financial return is indirect: fewer departures, less locum spending, better recruitment.
In the second version, the organisation notices the saved time within a quarter. Two extra appointment slots are added per day to improve access and justify the licence cost. Those two extra patients bring their own notes, orders, results and messages. The physician’s total workload is now slightly higher than before the scribe arrived, even though each individual note is easier. Burnout scores drift back up by the next annual survey. The scribe is judged to have “worked” on productivity and “not worked” on well-being.
There is a third version, which is probably the most common. Nobody makes an explicit decision. Schedules tighten a little as templates are revised, a few more same-day slots appear, and the inbox keeps growing. No one intended to recapture the time, but it is gone anyway.
Both versions are rational from someone’s point of view. Only one of them delivers what clinicians were promised.
The strongest counterargument
A reasonable health system executive would push back. Access is a real problem. Patients wait weeks for primary care appointments in many systems. If a technology frees up clinician time, using some of it to see more patients is not greed. It is meeting unmet need. And licences cost money that has to come from somewhere, often from budgets that could otherwise pay for more staff.
This argument partly holds. Keeping all of the saved time for clinicians while patients wait is hard to defend, particularly in publicly funded systems. And there is a risk that protecting time becomes a way of resisting any change in productivity at all.
The answer is not to forbid recapture but to be deliberate about it. An organisation can decide to split the gain, returning some time to clinicians and using some for access, and measure what happens to both burnout and waiting times. What undermines trust is recapture by stealth, where clinicians are sold a well-being tool and discover it was a throughput tool.
Second-order effects to watch
Several less obvious consequences are worth tracking.
Note inflation. AI drafts tend to be thorough. Thorough notes are longer. Longer notes are harder for the next clinician to read. If every note grows, the time saved in writing may be partly lost in reading.
Skills and judgement. Writing a note forces a clinician to organise their thinking. Some educators worry that trainees who never compose notes from scratch may lose a useful discipline. The evidence on this is thin, but it is a reasonable question for training programmes.
Patient experience. Many patients welcome a clinician who looks at them rather than the screen. Some will be uncomfortable being recorded, particularly in sensitive consultations such as mental health, sexual health or domestic violence. Clear opt-out processes matter.
Vendor concentration. As scribes become embedded in EHR platforms, the choice of tool may narrow to whatever the EHR vendor offers. That could simplify integration and reduce competition at the same time.
What this means for different players
Health system leaders should decide in advance how the benefits will be shared, communicate that to clinicians, and measure both burnout and access over at least a year. Setting a clear policy on schedule changes before rollout avoids the perception of a bait and switch.
Clinicians should keep reviewing drafts carefully. The note carries their signature and their liability. They should also report errors systematically, because vendors improve fastest when errors are documented, not just complained about.
Payers should watch coding patterns after adoption, but also consider whether documentation requirements themselves could be simplified. If notes can be generated automatically, some of the requirements that made them long in the first place may no longer serve their purpose.
Ministries and health systems outside the United States should notice that most evidence so far comes from US ambulatory care, with its particular billing and documentation culture. Where notes are shorter and billing is simpler, the gains may be smaller, and the case for investment should be tested locally.
The takeaway
Ambient AI scribes appear to deliver something rare in health technology: a tool clinicians actually like. The early burnout data are striking. But the benefit is fragile, because it rests on time and attention that the organisation will be tempted to reclaim.
The real test is not whether the note gets written faster. It is whether, a year from now, the clinician’s day is any shorter.
If your organisation has rolled out an ambient scribe, what happened to schedules and panel sizes afterwards? Tell us in the comments.
