The problem
Clinical documentation is one of medicine’s heaviest burdens. Kaiser Permanente adopted the technology “to respond to growing physician strain and burnout from the long hours they spend documenting patient care”, said Brian Hoberman, chief information officer of The Permanente Federation (The Permanente Federation).
The goal was twofold: cut the time spent writing notes, and free doctors from the keyboard while they are with patients (Kaiser Permanente Division of Research). Work done after hours, which physicians call “pajama time”, was a specific target (American Medical Association).
What they did
The tool records the conversation during a visit, with the patient’s approval, and produces a transcript and a draft summary that the physician edits before it enters the medical record. It does not make decisions or recommendations about care, and the audio recording is not kept (The Permanente Federation). Use was, and remains, voluntary for clinicians.
The Permanente Medical Group started with a 10-week pilot in late 2023, then changed vendors for the 2024 rollout; adoption kept rising with the new product (Kaiser Permanente Division of Research). Kaiser Permanente then ran a 10-week pilot of the enterprise tool in early 2024 before deploying it across its 8 regions, 600 medical offices and 40 hospitals.
Because of that scale, leaders commissioned a formal quality assurance plan. A team of physicians, informaticists, safety officers and evaluators reviewed star ratings, more than 3,600 free-text comments and a structured note-quality survey, and fed findings back to the vendor, which made rapid improvements (The Permanente Federation). The posts cited here do not name the vendors.
Results
According to the Kaiser Permanente Division of Research, 7,260 physicians used the scribe in 2,576,627 patient encounters between October 2023 and December 2024, and its use saved nearly 16,000 hours in documentation time compared with non-users (Kaiser Permanente Division of Research). The NEJM Catalyst analysis, written by the organization’s own researchers and reported by the AMA, puts the figure at 15,791 hours, equal to 1,794 eight-hour workdays (American Medical Association). Users had statistically significant reductions in note-taking time, time per appointment and after-hours work.
Benefits concentrated among heavy users. Some 3,447 physicians used the scribe in at least 100 encounters each (Kaiser Permanente Division of Research), and high users saved two and a half times more time per note than those who used it less often (The Permanente Federation). Adoption was highest in mental health, emergency medicine and primary care, departments with heavy documentation loads.
Surveys were positive. Among physicians, 82% said overall work satisfaction improved, according to The Permanente Federation. Among 118 patients surveyed, 47% thought their doctor spent less time looking at the computer, and two-thirds were comfortable with the technology (Kaiser Permanente Division of Research). By March 2025, the tool had supported more than 4 million encounters since the rollout across Kaiser Permanente (The Permanente Federation).
All of these figures come from Kaiser Permanente’s own researchers and publications; no outside body has audited them.
What didn’t go to plan
Note quality was good but not complete. In the quality assurance review, the lowest-rated domain was “note thoroughness”, and a minority of users said the tool struggled to track multiple speakers, left out information or made wrong assumptions that physicians then corrected (The Permanente Federation). Omissions matter beyond Kaiser Permanente: a January 2025 study that tested two commercial AI scribe products in a simulated clinical setting, reported by Medical Economics, found errors in 70% of the notes reviewed, with omissions the most common (Medical Economics).
Not every physician adopted it. Infrequent users cited a lack of integration with note templates and a sense that editing AI drafts took longer than typing from scratch, and users saw a small increase in time spent in the health record’s in-basket (American Medical Association). Infectious disease, obstetrics and gynecology, and urology physicians were the least likely to use it.
Cost is the open question. Kaiser Permanente has not disclosed what it pays. Medical Economics reports that enterprise platforms with deep integration often cost several hundred dollars per clinician per month.
Lessons
- Keep it voluntary and let adoption follow value. Use was voluntary, and adoption was highest in the departments with the heaviest documentation loads.
- Expect a dose response. High users saved two and a half times more time per note, so support for regular use matters more than wide licensing.
- Build quality assurance into the rollout. A dedicated team and structured feedback let the vendor fix issues quickly before full deployment.
- Watch for omissions, not just errors. “Note thoroughness” scored lowest, and missing details are the hardest mistakes for physicians to spot.
- Keep the physician as the decision-maker. Every draft is reviewed and edited before it enters the record.
Adapt it in your organization
Is this for you?
- Your professionals spend significant time documenting conversations, after hours included.
- You can measure time in your systems before and after, by user.
- Consent and recording rules for your setting are clear, and patients or clients can decline.
- A clinical or professional sponsor will own quality review.
- You can integrate drafts with existing templates and records.
A 90-day version
Pilot: offer an ambient scribe voluntarily to a few hundred clinicians in two high-documentation specialties, with structured note-quality ratings and system time logs compared against non-users. Owner: chief medical information officer. Success measure: after-hours documentation time down at least 10% among regular users, with note-quality ratings of at least 4 out of 5.
Metric to track
After-hours documentation time per clinician, split by usage level. Kaiser Permanente saw the largest gains among high users; its note-quality survey averaged 4.35 out of 5. For another case where automation stopped short of the last mile, see Airbnb’s test migration.
Failure modes to pre-empt
- Licensing everyone and measuring only averages, hiding that a minority of heavy users drive the gains.
- Treating draft notes as finished, so omissions reach the record.
- Ignoring template integration, which kept some physicians from adopting.
- Signing a multiyear contract without knowing cost per clinician.
Questions to take to your leadership team
- Which roles spend the most time documenting conversations, and how do we measure it today?
- Who reviews AI-drafted records for omissions, and how will we know if quality slips?
- What cost per user would the time saved justify?
Executive FAQ
How much time did Kaiser Permanente’s AI scribes save?
Nearly 16,000 hours of documentation time over 14 months for 7,260 physicians, compared with non-users, according to Kaiser Permanente’s Division of Research.
Do AI scribes make clinical decisions at Kaiser Permanente?
No. The tool drafts a visit summary that the physician reviews and edits; it makes no decisions or recommendations about care.
About the curator & AI disclosure
Curated by Luciano Ayres
Luciano Ayres is Director of Artificial Intelligence at ArcTouch, with more than 20 years of experience in software and technology leadership and specialized studies in deep learning. Connect with him on LinkedIn.
AI-generated briefing. This case was produced through an automated AI research and writing workflow curated by Luciano Ayres. Sources are linked for independent verification; summaries and analysis may contain errors and do not replace professional due diligence.