Some doctors say that their current best use for generative AI is to ease the heavy burden of documentation, which takes hours a day and contributes to burnout
The best use for generative A.I. in health care, doctors say, is to ease the heavy burden of documentation that takes them hours a day and contributes to burnout. Twitter: @zacharylipton , @stevelohr , @stevelohr , @chrissyfarr , and @morgancheatham Twitter: Zachary Lipton / @zacharylipton : Look mom, we made the paper! Exciting time for the team at @AbridgeHQ. https://twitter.com/... Steve Lohr / @stevelohr : “At this stage, we have to pick our use cases carefully. Reducing the documentation burden would be a huge win on its own.” - Dr. John Halamka of Mayo Clinic. https://www.nytimes.com/... Steve Lohr / @stevelohr : Generative artificial intelligence is coming to health care, but likely in measured steps. At first, more tireless scribe than genius partner. https://www.nytimes.com/... Christina Farr / @chrissyfarr : This piece hits most of the points but is missing some nuance (pun not intended) — uneven signing of BAAs — the complex regulatory environment — slower adoption amongst the non tech elite docs https://www.nytimes.com/... Morgan Cheatham / @morgancheatham : enjoyed reading this piece highlighting the real stories of physicians using @AbridgeHQ and the promise of generative AI in medicine “AI has allowed me, as a physician, to be 100 percent present for my patients.” https://www.nytimes.com/...
Context & Ripple Effects
This June 2023 piece captures the moment generative AI entered clinical medicine sideways: not through diagnosis but through paperwork. Dr. John Halamka of Mayo Clinic frames the discipline required — 'we have to pick our use cases carefully' — and the Abridge team's ambient-scribing work gets the nod as the early proof point.
The arc since confirms the thesis. Within a year, ambient clinical documentation had become its own product category, and by late 2024 doctors were routinely using tools like Microsoft's Nuance DAX Copilot to automate note-taking and summarization. Mayo Clinic's parallel work with Microsoft on AI assistants trained on its medical data shows the documentation wedge doubling as a data-collection strategy.
First-order effects
- Doctors gain back hours of daily charting time, directly attacking a named driver of burnout — the immediate payoff vendors like Abridge and Microsoft's Nuance are selling on.
- Hospitals evaluating generative AI get a low-risk starting point: documentation assistance touches notes, not diagnoses, sidestepping the clinical-error exposure that shadowed earlier medical AI.
Second-order effects
- EHR incumbents and rival scribing vendors are forced into a feature race around ambient documentation, since a tool that writes notes becomes table stakes in clinical software sales.
- Adoption pressure compounds from the malpractice side: Halamka's prediction that it may eventually be malpractice not to use AI turns documentation AI from optional perk toward expected standard of care.
Third-order effects
- If the pattern holds, medical AI commercialization settles into a two-track structure — administrative augmentation first, diagnostic support later — echoing the earlier warning that diagnostic systems like mammogram readers can worsen overtesting and overdiagnosis.
- Every recorded visit feeds training corpora: systems built on clinical data, like the Mayo Clinic–Microsoft model effort, turn documentation relief into a proprietary-data flywheel that concentrates advantage in a few health-system-vendor pairs.
The trend: Healthcare's generative AI adoption is sequencing administrative documentation before diagnosis, with ambient scribing as the wedge that also locks in clinical data access.