Practice position
AI assists. People decide.
Use approved systems, minimise information, check important output and keep accountable professional judgement human-led.
Begin with the service problem
Ask what problem the technology is intended to solve, for whom and compared with which non-AI alternative. A demonstration of fluent output is not evidence of improved access, quality, safety or productivity.
Expect governance from providers
Providers should be able to identify approved tools and uses, information controls, staff training, human oversight, incident routes, lived-experience involvement and periodic review. Unmanaged personal accounts may signal shadow AI rather than controlled innovation.
Evaluate claims in context
Request evidence relevant to the intended population and workflow. Define baseline, intended benefit, error types, inequality measures, staff burden and service-user experience. Include the possibility that the use should be changed or stopped.
Practical checklist
- Define the problem and alternative
- Ask what data is processed
- Name human accountability
- Require evidence proportionate to risk
- Agree monitoring, transparency and exit
Authoritative starting sources
NICE Evidence Standards Framework ↗NHS real-world AI evaluation lessons ↗Written by Tony D’Agostino / TD Consultancy. Reviewed . This page provides general professional education, not legal or clinical advice.