Frame the question
How can an organization evaluate an AI-enabled service before relying on it?
Look through several lenses
- Purpose and affected people
- Data origin and representativeness
- Human review and contestability
- Security, privacy and accessibility
- Monitoring and exit options
Worked learning case
Illustrative case: a library considers an automated tool that suggests reading material. Map the intended benefit, who may be left out, what staff can review, how a suggestion can be challenged and what happens if the tool is unavailable.
Trade-offs to compare
Consider feasibility, distribution of benefits and burdens, reversibility, time horizon, missing evidence and who should review the result.
Limits
This guide does not certify a system, replace a legal/privacy assessment or establish an IEU AI policy.