AI Scribes: 69% Less Charting, One New Habit

A study out of Toronto this month measured something every doctor feels: how much of a consult goes to the note. Without an AI scribe, physicians spent more than one-third of each encounter on note-taking. With one, documentation time fell by 69.1%.

That is a big number. It also comes with a catch the authors flag themselves.

What the study actually did

Nine physicians ran simulated primary care encounters, once with an AI scribe and once without. Researchers analysed video recordings to see where the time went. The work was published in JAMIA Open, led by University of Toronto PhD student LaShawn Murray, with Enid Montague and Dr. Onil Bhattacharyya of Women's College Hospital as co-authors.

Two limits matter. Nine doctors is a small group. And the encounters were simulated, not real clinics. The authors say more research is needed on everyday practice, and on how patients feel about being recorded.

So treat 69.1% as a signal, not a promise.

The detail worth your attention

Here is the part I keep thinking about. Some physicians picked up scratch pads while the scribe ran. They jotted their own clinical thoughts as the visit went on.

Murray described what she saw: physically creating the charts throughout the appointment prompted more thinking and more clinical decision-making questions from the doctors.

Read that twice. The typing was never only clerical. For some doctors, writing the note was also where the reasoning happened. Remove the typing and you remove a prompt to think. Some of them noticed, and rebuilt it with paper.

The other side of the scribe

Meanwhile, at U.Va. Health, the scribes are people. The student scribe program began in 2007 with five original scribes. DAX Copilot, an ambient tool that drafts notes from recorded conversations, was implemented there in 2025. The program's annual budget was cut by one-third, announced in April 2026. A new Emergency Department schedule began on September 15.

Students report the effect. One fourth-year used to work 16 to 20 hours nearly every week and is now scheduled for 20 hours across all of October. Another went from two or three eight-hour shifts a week to about one shift of four to five hours.

One physician quoted in the piece, Dr. Sarah Wendel, noted that every note still requires review, because DAX can add language or misinterpret statements.

A human scribe sat in the room and knew the patient. An ambient tool drafts from audio. The review step is now the safety net, and the doctor owns it.

What this means on a ward round

Ward rounds are not primary care visits. You walk between beds, you cover a list, you hold the whole team's questions in your head. A drafted note is useful, but it is one output of a long chain of thinking.

The Toronto result suggests where the next gain sits. If drafting the note gets cheaper, the scarce thing becomes the reasoning that comes before it, and the check that comes after. A scribe that only writes leaves both to you.

This is the bet behind Medows. The tool a doctor needs is not a faster stenographer. It is a workspace that holds the patient's context, lets you think in it, and shows where each line came from, so review takes seconds and not minutes. Documentation is part of the shift. It is not the shift.

A short checklist for skeptics

• Ask whether a time-saving figure comes from a real clinic or a simulation.
• Ask where the doctor's own thinking now happens.
• Ask who reviews the draft, and how easy the tool makes that review.

Sources

• AI scribes save doctors' time and may ease administrative burden, Medical Xpress (JAMIA Open study, DOI 10.1093/jamiaopen/ooag101)
• U.Va. Health cuts scribe hours as AI documentation tools expand across Health System, The Cavalier Daily