Health Systems Keep Clinicians in Loop, Resist Autonomous AI

Becker's Hospital Review reports that health systems have widely adopted AI for administrative tasks like IT triage, HR inquiries, coding, scheduling and claims denials, but are holding back on three categories: autonomous clinical agents, AI that independently diagnoses or orders treatment, and AI that interacts directly with patients. Executives from Mass General Brigham, UCHealth, Yale New Haven Health, Northwestern Medicine, Cooper University Health Care and CommonSpirit Health describe keeping a clinician in the loop as a deliberate governance choice driven by regulation, liability and unresolved risk of patient harm. Illinois has codified part of this boundary, barring AI from therapeutic decision-making in mental healthcare while allowing administrative uses. The piece notes some systems, like Hoag, have not deployed agentic AI at all, citing immature oversight infrastructure.