Using the cited guidance as a working standard, an AI use case is ready for evaluation when the team can show that the user’s problem is understood independently of a proposed solution and that the business value or context of the use is clearly defined. This gives evaluation a defined problem and purpose; it does not prove that the solution will work or deliver a specified result.
The GOV.UK Service Manual advises teams to “focus on the user’s problem rather than possible solutions.” NIST AI RMF states that “The business value or context of business use has been clearly defined.” Together, these statements support a simple distinction: the problem should be clear before the solution, while the intended business value, context, or both should be explicit.
How to check the evidence
| Dimension | Evidence to present | Diagnostic question |
|---|---|---|
| User problem | A concise description of the user’s problem, with the proposed solution removed or addressed separately. | Can a reader identify the problem without inferring it from the preferred solution? |
| Business value or context | A clear account of the intended business value, the context in which the use would occur, or both. | Can an evaluator identify the intended value or context without guessing? |
Each statement should also have inspectable supporting material. Describing a workflow before establishing the user problem, or calling an idea “high value” without explaining its relevance, does not by itself demonstrate clarity.
The table translates the two cited statements into a practical check; it is not a research method or evaluation framework prescribed by either source.
What the team must still confirm
Neither cited statement defines how much evidence is sufficient, which method should establish the user problem, or how business value and context should be documented. The team must separately determine:
- what supporting material is needed for the user-problem statement;
- how the business value or context will be evidenced;
- which measures, thresholds, and decision rules will govern evaluation;
- which assumptions and claims remain unverified.
The two clarity conditions provide a defensible starting point for evaluation. They do not replace project-specific evidence requirements or establish a successful outcome in advance.