AI writes the clinical note. Who teaches the review?

An October 12 commentary proposes training for clinicians who edit AI notes. It is a framework to test, rather than evidence that the training already works.

A commentary published in npj Digital Medicine on October 12, 2026 proposes teaching trainees how to review clinical notes drafted by ambient AI. Its authors, Yash Raka, Reed Van Deusen and Holly Wiberg, distinguish making documentation faster from preserving learning and checking accuracy.

Four skills, rather than a final click

The proposed framework covers finding errors, reconciling a note with the clinical reasoning of the encounter, writing independently and adjusting trust in the AI. The authors suggest simulation as a place to teach and evaluate these skills.

The proposal still needs testing

The authors explicitly call for empirical evaluation. This is a commentary, not a trial showing that its curriculum improves patient outcomes. Its starting context is U.S. medical education; applying it elsewhere would need adaptation.

Why the distinction matters

Our reading is that a workflow needs to define what its reviewer checks and how that ability is assessed. A person’s approval alone does not tell a reader which checks happened. This is an editorial implication, not a new measured result.

Published October 13, 2026. The source commentary was published October 12, 2026; no new experiment date is asserted. This article reports an education proposal, not patient care instructions.

Sources: npj Digital Medicine · review training.