If your staff don't have a sanctioned AI tool that actually helps them, they're already using one you don't know about.
A national survey by Wolters Kluwer Health found that roughly 40% of healthcare professionals have seen colleagues use AI tools their organization never approved. Nearly 20% admit they've used unapproved tools themselves (opens in a new tab). 10% have used them in direct patient care.
This isn't recklessness. It's staff drowning in documentation, inbox volume, and admin work reaching for whatever helps them get through the day.
But shadow AI carries risks that are almost invisible: confident-but-wrong answers, no audit trail, inconsistent reasoning, and patient information flowing into systems no one has vetted.
The fix isn't a stern memo about banned tools. It's giving staff a sanctioned option that's actually better than the unofficial one. Integrated into the workflow they already use. Traceable. Built for the work they actually do.
If the official tool is slower or worse than ChatGPT in a browser tab, the policy is already lost.
What this means for health organizations
The test for a sanctioned tool is practical: would staff choose it over the free alternative? If not, improving the tool will do more than writing a stricter policy.
