Back to Evidence & Guides

Clinic AI Self-Check

Twelve questions for one AI tool. Review the evidence your clinic holds and identify the questions that need follow-up.

Allow about twenty minutes. Complete the worksheet with the people responsible for the tool. Keep completed answers within your clinic.

  • Covers one named tool and use case.
  • Helps identify evidence gaps and assign actions.
  • Uses Yes, No, Not sure and N/A responses.
  • Does not establish safety, compliance or readiness.
  • Is separate from the existing two-minute Clinical AI Risk Check.

Get the worksheet

ELSA AI uses these details to send the worksheet and handle related communications. A download request is not permission for a marketing sequence. Privacy notice.

Question preview

Read-only. The worksheet is where answers are recorded. Response options are Yes, No, Not sure, N/A.

A. What is in use
  1. 1. Is there a written record of this tool, covering its purpose, who uses it and what information it handles?
  2. 2. Have staff been asked, through a confidential, role-level, non-disciplinary process, whether they use this or any similar tool through a personal account for clinic work?
B. Functions and change
  1. 3. Is there a written record of which functions are enabled, as deployed?
  2. 4. Is there a documented process to identify, review and authorise material changes to enabled AI functions?
C. Information and privacy
  1. 5. Where personal information is involved, are its flow, recipients, processing locations and any model-training or other reuse documented?
  2. 6. Is the actual retention and deletion arrangement documented for the information this tool handles?
  3. 7. Where personal information is involved, is DPIA screening documented and any required DPIA complete before use starts? For existing use, was it completed before processing began?
D. Supplier and contract
  1. 8. Are material supplier and clinic responsibilities documented, with processor terms in place where required and unresolved allocations recorded?
  2. 9. Are the clinic escalation route and supplier contact arrangements documented for errors, harmful outputs or information incidents?
E. Review, responsibility and patients
  1. 10. Are responsibilities assigned for approving changes to this tool's use and for responding to concerns about it?
  2. 11. For AI-drafted clinical content, are review criteria, source information, the approving clinician and sufficient review time defined?
  3. 12. Where AI affects patient care or information, are patient explanations and staff responses to questions or requests for alternatives documented?