AI appears in nearly every healthcare technology pitch now. The label tells you very little. Ask what the feature actually does, which records it reads, and who checks the result.
Ask for the exact task
The vendor should be able to name the input, output, user, and decision involved without falling back on “intelligent” or “automated.”
Ask where judgment remains human
Coding, documentation review, payer work, appeals, and patient communication can involve context and professional judgment. Clarify what the system suggests, what it performs, and what a person must verify.
Ask for product evidence
See the current capability in the product. Ask what happens when information is missing, confidence is low, or the output is wrong.
Ask about governance
Understand auditability, access, data handling, model or vendor dependencies, and how changes are evaluated.
Automation can save time when the task is specific and the result can be checked. A broad AI promise is not a substitute for showing the feature work.