A Philadelphia health system says doctors who used the tool spent less time on notes, but not every doctor chose it.
An AI that writes the doctor's notes
Jefferson Health, a hospital system in Philadelphia, has published a study of an AI tool that writes doctors' notes. The tool listens to a visit and drafts the note, which the doctor then checks. Jefferson says the clinicians who used it, meaning doctors and others who treat patients, spent less time on notes. They also did less charting after hours, a chore doctors call pajama time. The study came out on Oct. 23.
How many minutes came back
Jefferson compared the 12 weeks before and after each clinician started using the tool. The results cover 637 clinicians who used it often enough to count. They come in two groups, one for Jefferson and one for a partner it merged with, because the two used different record software. Time spent on notes per appointment fell 10.5% in one group and 17.3% in the other. After-hours charting fell by 8.9% and 16.6%. The quarter of clinicians who improved most cut their pajama time by nearly 17 minutes on each scheduled workday. That is the best case, not typical.
One physician told Luis Taveras, Jefferson's chief digital and information officer: “I go home. I can spend time with my family. I don’t have to go home and do my notes anymore.” Saved time can vanish into a fuller schedule. Jefferson says productivity held steady among users. A separate study of more than 8,500 clinicians at five academic medical centers linked an AI note tool to a 1.7% rise in weekly visits. None of the five centers required clinicians to book more patients to qualify for the tool. I read that as a small shift toward more visits. The Jefferson gains apply only to clinicians who chose the tool. How many did?
“I read the study as a modest gain for the clinicians who chose the tool, and I would not call it proven for everyone until clinicians are assigned to it by chance.”
Who chose the tool
Of 5,234 clinicians with access, 1,802 used the tool actively. That is about a third. Use was voluntary, and Taveras says growth will come from word of mouth, not a formal campaign. No one was assigned to a group by chance, as with a coin flip. So the study shows what happened to people who chose the tool. They may differ from the rest. The authors include Jefferson's president, Baligh Yehia, so the system partly graded its own tool. Still, Abridge, the company behind the tool, had no role in the study and gave no funding.
A stricter test comes from UCLA Health. There, researchers assigned 238 physicians by chance to use one of two AI note tools or to keep their usual routine. Doctors using Nabla cut their documentation time 9.5% more than the group that kept its usual routine did. Those using DAX saw a smaller drop that could have happened by chance.
What patients get
Jefferson says patient survey scores were the same or higher with the tool across more than 280,000 visits. That higher score mostly disappears when researchers compare a clinician's own visits, with and without the tool. Maybe the clinicians who chose it already had happier patients. The authors found no evidence that the tool harmed patient experience.
In the UCLA trial, fewer than 10% of patients declined the tool. Doctors said the AI drafts sometimes held errors that matter for care. One mild patient safety event was reported. UCLA internist Dr. John N. Mafi says doctors must review the drafts, not just accept them. Doctors and patients talk, and the tool listens. Jefferson is now trying it where people talk much less.
Nurses and quiet rooms
Jefferson's nursing pilot runs on three units at Abington Hospital. Roughly 30% of the nurses taking part are still hesitant to use it. Taveras says it is much harder there than with physicians. Most of the time, he says, nurses do not really talk: they take vitals, and there is a lot of silence.
I find that oddly reassuring: the software is no magic. It needs people talking, and someone willing to check its work. What I like is who benefited. The clinicians with the heaviest documentation loads got the biggest relief. Counting all daily documentation time, not only after hours, clinicians who are not physicians, such as nurse practitioners, got back 9.9 minutes. Physicians got back 3.8 minutes. Jefferson plans to make the nursing rollout opt-in, meaning nurses choose to join, as it did for physicians.
Nov 5, 2026
