How much time do AI scribes actually save?

The biggest number in a scribe pitch is two hours a day. A two-year JAMA study of 8,581 clinicians measured sixteen minutes. The gap between those two numbers is where your buying decision lives, so here is what the research found, and what to do with it.

How much time do AI scribes save?

Clinicians who adopted an ambient AI scribe spent 16 fewer minutes on documentation and 13.4 fewer minutes in the EHR overall for every eight scheduled hours of patient care. That figure comes from a multisite cohort study in JAMA, published in April 2026, which followed 8,581 ambulatory clinicians across five US academic health systems from June 2023 through August 2025. About 1,800 of them adopted a scribe; the rest became the comparison group.

Adopters also delivered 0.49 additional visits per week, roughly one extra visit every two weeks. The authors describe every one of these changes with the same adjective: modest.

The finding that deserves a spot on your whiteboard is the one that did not move. EHR time outside work hours did not change significantly. Clinicians got minutes back inside the workday, and the evening logins stayed right where they were. A scribe shortens the note; it does not, by itself, decide what the freed minutes get spent on. That takes a schedule decision from whoever runs the practice.

Why do the results vary so much?

Because the tool and the habit matter more than the category. A randomized trial at UCLA Health put two commercial scribes head to head: 238 physicians across 14 specialties were assigned to Nabla, to Microsoft’s DAX Copilot, or to a usual-care control for two months. Nabla cut time-in-note by 9.5 percent against control. DAX Copilot’s reduction was not statistically distinguishable from zero.

The absolute numbers are worth sitting with. Over the trial, control physicians trimmed 18 seconds off each note on their own, DAX users 23 seconds, Nabla users 41. Seconds per note, not hours per day. The trial was short and confined to one health system, and its authors flag both, but the spread between two polished, well-funded products is the point: “an AI scribe” is not one thing.

Usage intensity is the other lever. In the JAMA study, the clinicians who ran the scribe on 50 percent or more of their visits saw the biggest reductions, along with primary care clinicians, advanced practice clinicians, and female clinicians. The average includes plenty of clinicians who tried the tool and let it drift. The clinicians who made it the default are the ones the biggest reductions belonged to.

Do AI scribes reduce burnout?

AI scribes reduce burnout more reliably than they save time, with one caveat worth knowing. A study across six US health systems in JAMA Network Open surveyed 263 ambulatory physicians and advanced practice practitioners before and after scribe adoption in 2024. Burnout prevalence fell from 51.9 percent to 38.8 percent after 30 days of use. Participants also reported saving about 10.8 minutes per workday, a self-reported figure that lands in the same modest range the EHR logs do.

The caveat: it was a quality improvement study with no control group, built on self-report, and the sample likely leaned toward clinicians who wanted the technology to work. The UCLA trial, which did have a control group, points the same direction on its secondary endpoints. Physicians in the two scribe arms, analyzed together, reported lower task load and less work exhaustion, even though one of the two tools had barely moved the stopwatch.

That pattern is the honest heart of the evidence. Documentation feels dramatically lighter before it gets dramatically shorter. For an owner trying to keep good therapists, the feeling is not a soft outcome.

Does any of this apply to a PT clinic?

Apply it with eyes open, because the evidence so far comes entirely from physicians and advanced practice clinicians at large health systems, using general-purpose scribes integrated with hospital EHRs. Nobody has published a comparable trial in outpatient PT yet. The profession is engaging seriously in the meantime. APTA issued a practice advisory on ambient scribe technology in September 2025 and gave the tools a feature in APTA Magazine this August.

Two of the findings transfer cleanly to a PT schedule. First, the usage-intensity effect cuts in your favor: a therapy day is a dense stack of visits, and every one of them ends in a note. A scribe that runs on every visit compounds there in a way a physician’s clinic day never allows. Second, the same effect punishes friction. A scribe that drafts into its own app, waiting for someone to paste the draft into the chart, is exactly the tool that slides below 50 percent usage by month two. At that point you are paying full subscription price for the control group’s numbers.

That is the design bet behind Aurora, the ambient scribe built into Orion, whose whole brief is one line: “Focus on your patient, not your keyboard.” The draft lands inside the visit it belongs to as a PT-specific SOAP note, and review and sign is the last tap of the appointment. It is included on both plans, with no per-note metering. The workflow question, where the note lands and who moves it, is the one we would tell you to press any vendor on, ours included. Our guide to choosing an AI scribe for a PT practice walks through that vetting, and the companion answer on how accurate AI scribes are covers the review-before-signing side.

What should a practice owner actually expect?

Plan on the JAMA numbers, not the pitch-deck ones: minutes per day, not hours, arriving only for the therapists who use the tool on essentially every visit. Then decide in advance where those minutes go, because the five-system data says evenings do not shrink on their own. A practice that books nothing into the recovered time gets a calmer 3pm. A practice that moves one documentation block into patient care gets the 0.49 extra visits, which across a few therapists and fifty weeks quietly pays for a lot of software.

And when you trial one, run it against your own stopwatch on your ugliest week, the way the trials did. The clearest lesson in all of this research is not the sixteen minutes. It is the gap between the clinicians who signed up and the clinicians who kept using it. Buy for month three, not for the demo.