Summary
Two hours of documentation and desk work for every hour of direct patient care. Measured in 2016, unmoved since.1
Our industry has spent nine years and a great deal of money on that number and the best randomised result is 41 seconds.2 We were part of that and we think it was the wrong target.
The briefing, the letters, the coding and the phones. Built for the forty minutes, not the ninety seconds.
Book a demoThe number
Sinsky and colleagues followed 57 physicians across four specialties with trained observers. Nearly two hours on the record and desk work for every hour with patients, plus one to two more hours at home most nights.11Sinsky C, Colligan L, Li L, et al. Allocation of physician time in ambulatory practice. Ann Intern Med 2016;165:753-760. Published 2016.
Everything since has been built to attack it. The number has not moved.
What we all built
Microphones. Every serious vendor in this category, including us, shipped something that listens to a room and writes the note. UCLA randomised 238 physicians across roughly 72,000 encounters to test two of them. Nabla cut note-writing time 9.5 percent. DAX Copilot showed no statistically significant change. Neither reduced after-hours record time.22Lukac PJ, Turner W, Vangala S, et al. Ambient AI Scribes in Clinical Practice, A Randomized Trial. NEJM AI 2025;2(12). 238 physicians, 14 specialties, approximately 72,000 encounters.
Forty-one seconds, against two hours. The trial is not an indictment of those products. It is a measurement of how much of the two hours a microphone can reach.
What the two hours is made of
Open the list at six. Letters. Results that came back while you were consulting. Messages. Prior authorisations. Forms. Callbacks you promised. Coding queries from billing.
The note is one line. Every product built for this problem reaches that one line, and the arithmetic of 9.5 percent of one line is 41 seconds. The trial and the list agree.
Why we all aimed there
A microphone demos. You put it on a stage and text appears and the room reacts. Nobody has ever demoed a prior authorisation matrix.
And the note was the only part sitting inside the consultation, and a listening product reaches nowhere else. Everything else on that list is read out of a record, not heard in a room. So the category built what its architecture allowed and called it the problem.
What we missed while we were counting seconds
From 2026 your E/M level turns on Medical Decision Making, so the note is the evidence for the code.55From 2026, Medical Decision Making is the primary basis for E/M level selection. AAPC reports up to 19 percent of E/M visits undercoded at roughly $37 each. Reported MA downcoding runs $42 to $74 per visit. Up to 19 percent of E/M visits are undercoded at roughly $37 each, and Medicare Advantage downcoding takes $42 to $74 off a visit and never appears as a denial. See our essay on that.
What we did about it
We built for the list rather than the room. The briefing assembled from the chart before the patient sits down. The letters drafted from the consultation and the record together. The coding attached at the point of decision. The phone answered by something that can see what happened.
What we are not claiming
We have not moved the two hours either and we have no trial that says we have. Nabla has 41 randomised seconds, which is 41 more than we can evidence.
Sinsky is nine years old. Practice has changed, EHRs have changed, and nobody has repeated it at that quality, which you should find as odd as we do.
Availability
WA\ Clinician runs a 14-day free trial with no minimum term. Run it and look at your list at six. Pricing is on one page.
Frequently asked questions
Has the two-hour documentation burden changed since 2016?
No. Sinsky and colleagues followed 57 physicians across four specialties with trained observers and found nearly two hours on the record and desk work for every hour with patients, plus one to two more at home most nights. Published 2016. Everything since has been built to attack it and the number has not moved. The best randomised result any ambient scribe has produced is 41 seconds a note, and neither product tested reduced after-hours record time at all.
Why have AI scribes not solved the documentation burden?
Because the note is one line on a long list. Open the list at six and it is letters, results that came back while you were consulting, messages, prior authorisations, forms, callbacks and coding queries. A microphone reaches the note and nothing else, because everything else is read out of a record rather than heard in a room. Nine and a half percent of one line is 41 seconds. The trial and the list agree.
Why did AI scribe companies focus on the consultation?
Two reasons, and one of them is honest. The note was the only part of the problem sitting inside the consultation, and a listening product reaches nowhere else, so the category built what its architecture allowed. The other reason is that a microphone demos. You put it on a stage and text appears and the room reacts. Nobody has ever demoed a prior authorisation matrix to applause.
Bring us the forty minutes.
Thirty minutes, your own workflow, running live. Bring the work that is still open at six o'clock.
About this article. Written and published by WA\, which sells clinical AI and is arguing that its own category aimed at the wrong target, including us. The time-motion research and the randomised trial cited are independent peer-reviewed work. The Sinsky data is from 2016 and has not been repeated at equivalent quality. WA\ has no peer-reviewed clinical trials published to date and does not claim any. Coding figures are third-party analysis.