The Procurement Self-Test

You are not scoring the tool. You are scoring what you can prove about it.

My full evaluation instrument scores clinical AI tools across eight domains, with red-flag caps and a cross-domain safety interlock. That instrument stays with me. What follows is the buyer-side mirror of it: one question per domain, asking whether you currently hold the evidence each domain needs.

Answer for one specific tool you are being sold, or already own. Be honest. A vendor slide is not evidence you hold; it is evidence they showed you.

  1. Domain 1 · Clinical Evidence QualityCan you name a peer-reviewed study of this tool, on the version you are actually buying, that was not authored or funded solely by the vendor?

  2. Domain 2 · Population ValidityHas the tool been validated on a population that matches yours, in demographics, language and case mix, with results you have personally seen?

  3. Domain 3 · Operational PerformanceDo you know the tool's real-world alert burden and false alarm rate at a live site comparable to yours, as distinct from its headline accuracy figure?

  4. Domain 4 · Workflow IntegrationCan you describe exactly where this tool sits in your clinicians' existing workflow, and what a clinician must do differently at the moment it fires?

  5. Domain 5 · Regulatory ComplianceDo you know the tool's regulatory status in your jurisdiction specifically, including whether your regulator requires local validation before deployment?

  6. Domain 6 · Data Governance and SovereigntyDo you know where your patients' data will physically reside, who can access it, and what the vendor may retain or reuse, including for model training?

  7. Domain 7 · Implementation MaturityHas the vendor given you a named reference site of similar size and system to yours, which you are free to contact directly and privately?

  8. Domain 8 · Ethical and Cultural CompatibilityHas anyone assessed the tool's outputs, defaults and design against your population's ethical and cultural requirements? For Gulf systems this includes Islamic bioethics compatibility.

Nothing you have entered leaves this page. There is no tracking on this site.

This self-test is an educational instrument. It reports the completeness of your evidence file as you have described it. It does not evaluate any vendor or product, it is not procurement advice, and it is not a substitute for structured evaluation. The full instrument, its scoring logic, red-flag caps and safety interlock are not reproduced here.

For context on why composite scorecards miss deal-killers, read Why Clinical AI Procurement Scorecards Fail.