On the day an AI tool goes live, the work changes rather than ends. The tool now runs on real patients, inside workflows it can disrupt, on data that will drift away from what it was trained on. Someone has to check that it works on your patients, watch it, log what it does, train the people who use it and answer for it when something goes wrong. The chapters below cover each of those tasks.
FUTURE-AI Lekadir et al. (2025) places these recommendations in the deployment phase: Un4 (show that the tool is clinically valid locally), T3 (check the quality of inputs and outputs), T4 (audit and update the tool periodically), T5 (log how the tool is used), Us3 (provide training), G5 (identify and meet regulatory requirements) and T6 (set up AI governance).
Chapters in this section¶
| Chapter | What you will learn |
|---|---|
| Local validation | How to check that the tool works at your site before go-live |
| Integration into clinical workflows | The technical and organisational work of fitting the tool into care |
| Quality control | How to catch bad inputs and implausible outputs while the tool runs |
| Monitoring and auditing | How to detect a drop in performance, run audits and act on them |
| Logging and traceability | What to log, how long to keep it and how to use it |
| Training and onboarding | How to prepare staff to use the tool safely, including the AI literacy duty |
| Regulatory compliance | Obligations after go-live under the MDR/IVDR and the EU AI Act |
| Governance and accountability | Who owns the tool, who decides about changes and how patients can raise concerns |
- Lekadir, K., Frangi, A. F., Porras, A. R., & others. (2025). FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare. BMJ, 388, e081554. 10.1136/bmj-2024-081554