Ambient AI scribes are sold on a generous promise: give clinicians their time back. The early evidence backs it. In a six-system study, clinician burnout fell from 51.9% to 38.8% within a month of adoption, according to a 2025 article published in JAMA Network Open. Anyone who has finished a clinic day with two hours of notes still waiting knows why that number matters.
But health systems are not charities for clinician well-being. When a system buys an ambient scribe, what is it actually buying?
The conversion assumption
The industry’s own marketing answers. An implementation playbook published by a wireless-health industry association puts it plainly: “Most vendors cite minutes saved per visit, but ROI only materializes when those minutes are converted into additional capacity, revenue, or access, or when they demonstrably reduce clinician burnout costs,” said a blog post from a 2026 industry blog post by Wireless-Life Sciences Alliance. A return appears only when the minutes are converted into more visits, into richer coding, or into a measurable drop in what burnout costs the organization.
That last channel is the one clinicians are counting on, and it is the hardest to measure. Burnout’s cost is real, about $7,600 per employed physician per year, driven by turnover and reduced clinical hours, according to a 2019 study published in Annals of Internal Medicine. But its impact is diffuse, delayed, and hard to directly attribute. In contrast, more visits and higher coding are easy to calculate. As a result, a finance office has clear incentive to default to the quantitative options.
The experiment has already been run
Human scribes spent 20 years making documentation more efficient inside fee-for-service payment. A meta-analysis of 39 studies found throughput rose by about 0.3 patients per hour and RVUs by 0.55 per hour, according to a 2021 study published in Annals of Emergency Medicine. The saved time did not, on average, stay with the clinician. It became volume, because the payment system rewarded volume.
Efficiency that meets a volume incentive gets absorbed as volume. The pattern is old enough to have a name, the Jevons paradox, and it has already been mapped onto health AI, according to a 2025 study published in The Lancet Digital Health. Today’s saved 30 minutes becomes the baseline tomorrow’s schedule is built on. The target resets to the higher number.
The quieter channel: coding, not just volume
While volume is the obvious channel, it may not be the largest one. A scribe that captures more of what happened in a visit also captures more billable detail, raising payment for the same encounter. The early data fit this pattern better than they fit a volume story: Riverside Health in Virginia recorded an 11 percent rise in work RVUs and a 14 percent rise in documented hierarchical condition category diagnoses per encounter after adoption, in a system-reported analysis published by the scribe’s vendor. A 2025 policy brief in npj Digital Medicine now warns of a coming “coding arms race.”
Some of this complexity was always in the room and never fully documented due to operational pressure. Paying more is the legitimate purpose of risk adjustment, and it can route resources toward sicker panels. However, more complete documentation slides easily into higher coding for its own sake, and from the outside no one can yet tell the difference. The UCSF authors could not: they flagged that their measured RVU gains might be more services, more accurate coding, or upcoding. Payers are already recalibrating in response.
The gap the buyer will want closed
What to do with potential time saved depends in part on the magnitude of the savings. The industry playbook’s own worked example assumes that “saving three to five minutes per visit across a 20-visit day yields 60 to 100 minutes of reclaimed time,” according to a 2026 article published by Wireless-Life Sciences Alliance. On the coding side, a major vendor advertises about $13,000 in incremental annual revenue per clinician from more complete diagnosis and visit-level coding, in a deployment validated by KLAS Research.
The strongest measurement to date is a trickle by comparison. A UCSF study published in JAMA Network Open of 1,565 physicians across 1.2 million encounters found an association of about 1.81 RVUs per week among adopters, which the authors value at roughly $3,044 per physician per year using the 2025 Medicare Physician Fee Schedule.
Set that against the $13,000 a vendor advertises and the gap is more than fourfold. The time side is wider still. A 2026 study published in JAMA of five academic centers found documentation time down about 16 minutes per day, against the 60 to 100 minutes the playbook’s example assumes, with visit volume up about half a visit per week, one extra patient every two weeks.
The discrepancy between industry numbers and real-world estimates may still reflect early adoption: only about 1 in 3 clinicians in the multisite study used the tool heavily, and the heavy users saved two to three times as much time. It is also possible the advertised numbers were never realistic. Regardless, a system that paid expecting $13,000 and is seeing $3,044 has every reason to chase the difference. Once the vendor has been paid, closing the gap is the buyer’s problem. Unfortunately the easiest solution leads directly to the people using the tool.
Where the ratchet bites
The effect will not be uniform. It bites hardest where fee-for-service payment meets employed clinicians on productivity targets in documentation-heavy specialties. Each condition supplies a different part of the mechanism. Fee-for-service is what makes a reclaimed minute convertible into revenue in the first place. Employment on a productivity target is what moves the decision out of the clinician’s hands, because the person who rebuilds the schedule is not the person who saved the time. And documentation-heavy specialties are where enough minutes exist to be worth taking, which is why the multisite study found its largest changes among primary care clinicians and the heaviest users. Remove any one of the three and the conversion weakens: the minutes are not worth money, or not anyone else’s to reassign or not numerous enough to bother with. It is muted, even reversed, under capitation or salary, where an added visit is a cost and freed time is likelier to stay freed.
That fee-for-service elicits more visits than capitation among primary care physicians is long established, per a 2000 review of four studies published in Cochrane Database of Systematic Reviews, whose authors themselves caution about generalizability. Whether a clinician keeps the relief depends less on the scribe than on how the clinician’s employer is paid.
So who gets the time back?
This is the conversation worth having now, while the tool is new and the norms are soft. Whether ambient AI relieves clinician burnout or accelerates clinicians to a higher RVU baseline will be settled in contracts and budget meetings, not in model benchmarks. Questions worth asking before the targets reset:
- Is any portion of the saved time protected, contractually, as time rather than capacity?
- Are volume and coding targets being re-baselined to the post-scribe number, and on what evidence?
- Is burnout being remeasured after targets reset, not just in the honeymoon month?
- Under value-based arrangements, can the efficiency be banked as access for patients who lack it, rather than as throughput?
The honeymoon data are in, and they are good. The number that will matter is burnout the quarter after the targets reset. So far, no one is scheduled to measure it.
Dr. Mathews is a practicing gastroenterologist and chair of the American Gastroenterological Association’s Committee for GI Innovation & Technology. He was previously head of clinical innovation at the Johns Hopkins Armstrong Institute for Patient Safety and Quality and a chief medical officer in digital health. Views are his own.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
