A study of more than 1,500 clinicians at Renton, Wash.-based Providence found that ambient AI documentation tools reduced note-writing time and after-hours documentation burden, with modest per-provider gains that compounded into system-level relief at scale.
Published May 29 in JAMA Network Open, the study is among the largest real-world evaluations of ambient AI at a nonacademic health system. Researchers drew on objective EHR metadata — not self-reported surveys — to measure change, tracking 16,149 observation-months from 1,547 active users between July 2023 and March 2025. Most participants were physicians; nearly two-thirds practiced in primary care.
Clinicians cut note-writing time almost immediately after adoption. After-hours documentation, known as pajama time, declined more gradually but showed a sustained monthly decrease throughout the study period. Providers also posted a small but statistically significant rise in relative value units — without taking on additional appointments — suggesting productivity gains came from recovered time rather than a heavier schedule.
“These findings align closely with the day-to-day experience of practicing clinicians,” said Scott Smitherman, MD, associate vice president and chief medical information officer at Providence Clinical Network, in a May 29 news release. “From a physician perspective, ambient AI isn’t about seeing more patients — it’s about reducing cognitive load and reclaiming time spent on administrative work so we can focus more fully on patient care.”
The researchers caution that per-provider improvements were modest and that voluntary adoption reached only about 8% of eligible clinicians by the study’s close. Providence serves more than 2 million patients across seven states, which means even incremental per-clinician gains aggregate into a meaningful organizational effect.
