9 fixes for the AI your staff is already using

A scheduler has a difficult message to send a patient, so she drops the patient's name, appointment history and a note about a missed payment into a free chatbot and asks it to soften the tone. The reply is good. She sends it. Nothing about the exchange is logged, reviewed or covered by a contract, and no one else in the practice knows it happened.

That is shadow AI, and it is not rare. Physician use of AI has more than doubled since 2023, with 81 percent now using it professionally in the AMA's 2026 Physician Survey on Augmented Intelligence. A Wolters Kluwer survey of health care professionals found 40 percent had encountered an unauthorized AI tool at their organization and another 17 percent admitted using one. Meanwhile HHS now names a third-party AI chatbot handling patient information as a business associate, which means an agreement is required before any of that data moves.

Banning the tools outright pushes the behavior further underground. The practices that get this right make the sanctioned path faster than the sneaky one. Here are nine places to start.