Hospital AI oversight lagging; senior leaders must take daily responsibility

AI tools are now common in U.S. hospitals, handling tasks such as drafting clinical notes, flagging sepsis, screening imaging, processing prior authorizations, and responding to patient messages.

While the benefits are evident, most health systems still monitor these tools with the same slow, checklist-driven process used for older equipment like MRI scanners, often requiring quarterly meetings and six-month approvals.

The authors argue that this model is unsafe for AI and that fixing problems should not be left to IT departments alone; senior leadership must assume direct responsibility for AI oversight.

They call for daily AI governance to become a routine part of executive duties, ensuring that safety and performance are continuously monitored rather than reviewed only intermittently.