AI-generated replies are already sitting inside physician inboxes, promising to reduce one of the fastest-growing sources of administrative work.
A new NEJM AI study examined what happened after physicians opened those drafts. Across more than 14,000 AI-assisted responses, clinicians frequently changed scheduling instructions, lifestyle advice, diagnostic explanations, and test-result interpretations.
The clinically complex edits took the longest per message. Yet the largest cumulative burden came from a much more routine task: scheduling.
The findings give health systems a practical map of where AI messaging still creates work—and where a focused AI agent might actually remove it.
👉 Full breakdown in my latest Huddle+ article here.
