An ambient AI tool has been live in your clinic for a year. The notes largely draft themselves, the providers who use it say they would not give it back, and yet nothing about 9pm has changed. Charts still get closed from the kitchen table, the inbox is still deep on Monday morning, and the coding queue still needs a person to work it. That gap, between a shorter note and an unchanged evening, is the case for something wider than drafting, and it is where a medical scribe from Scribe-X is positioned. The gap is now measured rather than asserted: across 8,581 ambulatory clinicians at five academic health systems, documentation time fell by 16 minutes a day while time in the electronic health record outside working hours did not move significantly at all.
A five-system study found the evenings unchanged
The figures come from a multisite study published in JAMA in April 2026, which tracked adoption at Mass General Brigham, Emory Healthcare, UC San Francisco, Yale New Haven Health and UC Davis between June 2023 and August 2025. Of the 8,581 clinicians followed, 1,809 adopted an AI scribe.
Adoption was associated with 16.0 fewer minutes of documentation time per day, a relative fall of about 10 percent, and 13.4 fewer minutes of total EHR time, a relative fall of about 3 percent. Weekly visit volume rose by 0.49 visits. Record time outside scheduled hours showed no significant change. The revenue effect of the extra visits came to 167 dollars per month per adopting clinician.
Drafting works, in other words. The evenings the tools were bought to fix held steady.
So where does the time actually go?
Drafting addresses one slice of a much larger clerical load. The time and motion study run by the American Medical Association with Dartmouth-Hitchcock, published in Annals of Internal Medicine in September 2016, found physicians spending 27 percent of the office day in direct clinical face time with patients and 49 percent on the record and desk work, with a further one to two hours of clerical work at night.
Almost none of that second figure is note prose. It is chart prep before the patient walks in, problem and medication review, orders and referrals and prescriptions staged during the encounter, results and messages afterwards, then the coding and quality work that decides whether the visit is paid at the level of service actually delivered. An ambient tool that ends at the signed note hands all of it back.
The half-finished chart
The adoption figures in the JAMA study are the part worth reading twice. Only 32 percent of clinicians with a scribe available used it in more than half of their visits. Those heavier users saw roughly twice the reduction in total EHR time and three times the reduction in documentation time reported for the group as a whole.
The benefit concentrates, then, in providers who already drive their own technology. The provider who will not learn a new workflow, the locum covering two sessions a week, the specialist whose visits do not fit any template: they generate the same drafts and finish the same charts as before, and they are frequently the providers a practice is most worried about losing. A draft that still needs a provider to complete it is still work waiting on a provider.
Coverage has to be set provider by provider
That variation is why the more recent offers are built as a mix rather than as a single product. Scribe-X sets the level per provider across three solutions. On Solo, automation drafts the note at encounter close and the provider reviews and finalises it on their own schedule, which suits teams staffed to keep the finish in house. On Assist, a trained Remote Healthcare Assistant completes the note into the EHR, carrying entry, orders and codes. On Live, a trained Remote Medical Scribe documents the encounter in real time, so the chart closes at the visit.
Two operational details matter more than the tier names. Most solutions require no EHR integration, so IT is not asked to carry a project, and a business associate agreement is executed before any provider encounter. Both questions are worth asking in the first conversation, because the answers decide how much of the implementation lands on your own team.
What should a practice measure before it commits?
Documentation time is the wrong headline metric, because it is the one number every drafting tool moves. After-hours EHR time is the honest one, since it is precisely what the JAMA cohort failed to shift. Days to close carries the revenue cycle story. The distribution of level-of-service codes shows whether the record supports the care given. Retention shows whether any of it reached the people you were trying to keep.
Vendors publish their own ranges, and they should be read as vendor data. Scribe-X reports, from its 2025 client cohort, 1.9 to 3.8 more patients per provider per day, 28 to 62 percent faster days to close, and a 20 to 40 percent improvement in provider satisfaction. Those are wide bands, and the width is informative: what a practice gets depends on which providers sit on which solution. Ask for the baseline and the measurement window behind any figure of that kind, then run the same measurement in your own clinic, before and after.
