Yes—but the team should first clarify the user problem, current process, and intended value, rather than treating a proposed AI feature as the solution. Starting with the solution can conceal weak assumptions about what actually needs to change.
How to check the workflow before building
-
Start with the user’s problem.
The GOV.UK Service Manual advises focusing on “the user’s problem rather than possible solutions.” The team should therefore be able to describe what users are trying to accomplish, where the current workflow fails them, and what evidence supports that problem—not merely which AI capability seems relevant. -
Define the business value and context.
NIST AI RMF says that the business value or context of AI use should be clearly defined. Before development begins, the team should state what business activity the feature would support, what value it is expected to address, and which contextual conditions matter. Assumptions should be identified as assumptions rather than presented as established facts. -
Compare the proposal with the existing workflow.
Check whether the problem can be addressed by changing the process itself, including its steps, rules, handoffs, or interface. This comparison follows from a problem-first approach; it is not a requirement established by either cited source. If a simpler workflow change addresses the problem, that should remain the leading option. -
Make the proposed feature testable.
Describe the current workflow, the intended change, and the evidence that would indicate whether the feature is useful. The cited guidance does not provide universal success metrics, so the team must choose measures that reflect its own user problem and defined business value.
What the team must still confirm
The cited guidance does not establish that every team must redesign its workflow before conducting any AI experiment. It also does not show that AI is necessary for a particular use case.
Before committing to development, the team must confirm:
- that the user problem is supported by evidence;
- that the business value and operating context are understood;
- that the current process has been examined rather than assumed;
- that the proposed feature addresses a need more appropriately than simpler alternatives; and
- that the team can evaluate the feature against its stated purpose.
The practical answer is therefore conditional: improve or clarify the workflow first when the user problem or business value remains unclear. Once those elements are defined, an AI feature may proceed to scoped evaluation—but it should not be treated as automatically necessary or guaranteed to deliver value.