AI scribes hit the ‘modest savings’ chapter—humans still write better notes
Industry Buzz
Although ambient AI scribes hold promise for reducing clinician burden, independent, vendor-neutral evaluations of note quality are essential before large-scale clinical deployment.
—Study authors, via Annals of Internal Medicine
AI scribe vendors have for years promised a future where physicians spend less time documenting and more time caring for patients. That future may be arriving—but perhaps not as dramatically as advertised.
Taken together, the findings point to a new normal for primary care. AI scribes may be gaining a foothold in clinical practice, but their value depends on consistent oversight, careful review, and ongoing monitoring.
The efficiency gains are real, but smaller than expected
A large multicenter study published in JAMA examined EHR data from 8,581 clinicians, including attending physicians, advanced practice providers, and residents across five academic health systems. Among the 1,809 clinicians using ambient AI scribes, researchers found an average reduction of approximately 16 minutes of documentation time and 13.4 minutes of total EHR time per 8-hour shift compared to nonusers.[]
The benefits were more pronounced among PCPs and those who used AI scribes during at least half of patient encounters. There was also a modest increase in productivity, equivalent to roughly half an additional patient visit per week.[]
While those findings are encouraging, they fall short of the dramatic time savings often cited in marketing materials. Notably, AI scribe use was not associated with significant reductions of after-hours EHR work, suggesting that clinicians may still be spending considerable time reviewing and refining AI-generated documentation.[]
Better workflow does not necessarily mean better notes
In a study published in Annals of Internal Medicine, researchers compared notes generated by 11 ambient AI scribe platforms with notes written by 18 human clinicians using audio recordings from standardized primary care visits.[]
Human-generated notes consistently outperformed AI-generated notes across every measured domain, including accuracy, thoroughness, usefulness, comprehensiveness, organization, and conciseness.[]
While AI documentation tools are highly effective at capturing conversations, they are less reliable at interpreting clinical nuance, prioritizing information, and constructing a polished clinical narrative.
The study’s authors concluded that AI-generated notes should be considered drafts requiring clinician review rather than finalized documentation.[]
“Although ambient AI scribes hold promise for reducing clinician burden, independent, vendor-neutral evaluations of note quality are essential before large-scale clinical deployment,” the study’s authors note.[]
Adoption should be paired with auditing
For busy PCPs, AI scribes can help reduce administrative burden, but they do not eliminate the need for physician oversight. As adoption accelerates, clinicians may need to develop new habits around auditing note quality, identifying omissions, and monitoring recurring errors.
The technology appears increasingly capable of saving time. But for now, the responsibility for ensuring the accuracy of the medical record remains firmly human.