The method
Clinician Sign
Five questions, asked before an AI tool touches a patient.
Every product, process and policy I review gets the same test: would a real clinician sign this? If the answer is no, the problem is not the clinician.
Why it exists
AI does not fail loudly. It fails confidently. A system that cannot ground an answer produces something plausible, well-formatted and wrong, in the same tone it uses when it is right. That is what makes it dangerous to a clinical team: it reads as trustworthy.
The review work then lands on the same nurses who are already too busy to do everything. A tool is not safe because a vendor says it is. It is safe when a credentialed clinician can look at it, understand where it breaks, and put their name on it.
Clinician Sign is how I do that review. It is not a product and it does not need to be bought. It is a sequence of questions, and anyone evaluating clinical AI can use them.
The questions, in order
1. What does it do when it can't ground an answer in the source data?
This is the important version of "what happens when it isn't sure". In most deployed systems you get a fluent sentence either way, and the confidence signal a developer sees in testing is often not surfaced to the user at all. So don't ask to see the low-confidence path. Ask them to show you a case where it declined to answer. If they can't produce one, it doesn't decline.
2. Where does the output go when nobody checks it?
Trace the whole flow, because the risk isn't where you'd expect. Signed notes require authentication, so a drafted note reaching the chart unreviewed is a narrow path. The wider one is auto-populated flowsheet and vitals fields, data that flows into the record without a person ever affirming it. Find out what the tool writes and who has to approve it, field by field.
3. What is it actually reading?
Ask what data source it draws on and, more usefully, what it ignores. A tool that can't see the medication list is making recommendations with a blind spot. It won't tell you which one.
4. Who is accountable when it's wrong?
Ask the vendor to put it in writing. You'll usually get a version of "the clinician". Get them to say which clinician, for what decision, and under what conditions, because that answer determines how much the tool should be doing unsupervised.
5. Can you show me the misses?
Any vendor can demo a system being right. Ask what a false negative looks like, and how often one has happened in the field. The quality of that answer tells you more than any accuracy figure on a slide.
Where it applies
Fifteen years of clinical practice across emergency, oncology and infusion, plus building AI systems myself. That combination is the point: the review is done by someone who has carried a pager and shipped the automation.
Clinical-safety and evaluation review
For AI-health products. I sit on the clinical-safety and eval story until a credentialed nurse could sign it.
$200 / hourOperations audit
Capacity, staffing model, AKS/Stark, and a 90-day plan. I run a clinical floor. I will audit yours.
$5–15k / engagementIndependence
No vendor ties. I do not resell, partner with, or take referral fees from any product I review. That is the whole reason the sign is worth anything. An evaluation from someone with a product to protect is a sales document with a clinical font.
This is an independent advisory practice. Views on this site are mine, and reviewing a product is not an endorsement of it by anyone I work with.