You ask for three solutions to a recurring admin problem. The assistant suggests three versions of buying software. The ranking looks useful, but the choice was narrowed before anyone examined the problem.
Use AI first to generate genuinely different options, then evaluate them against agreed criteria. Separate the two tasks so an early favourite does not determine the whole list. Keep the options connected to a defined operating problem and realistic constraints.
Photo by Jo Szczepanska on Unsplash.
Describe the problem without a preferred solution
State what happens, who is affected and what an improved result would look like. Avoid naming a tool unless the task is explicitly about that tool.
For an illustrative onboarding issue, the problem may be repeated clarification about missing documents. Options could concern the request wording, timing, ownership or a supporting tool. The initial brief should leave room to consider each mechanism.
Explain constraints such as existing systems, available staff time and activities that cannot change. Distinguish a firm boundary from a preference the team could reconsider.
Ask for contrasting approaches
Use an option card with:
Mechanism: how this changes the work.
What remains manual.
Required business decisions.
Dependencies and information needed.
Main trade-off.
Situation in which it would be unsuitable.
Request differences in mechanism rather than differences in branding. Two products performing the same step may belong under one approach until the operating choice is clearer.
An illustrative option set
For missing onboarding information, one option is simplifying the request. Another is adding an early completeness check with a named owner. A third is changing the point at which the request is sent.
These could be combined later, but first ask what each would solve. Software may support an option, yet it is not the only possible intervention.
Ask the assistant to identify what evidence would distinguish the alternatives. A suitable next step may be observing a few current cases or discussing the rule with the process owner.
Evaluate after the list exists
Agree criteria such as expected usefulness, effort, dependencies and reversibility. Do not ask the tool to invent numerical scores as though they were measured facts.
Use qualitative reasoning where evidence is limited. Record what would need to be true for an option to work and what could invalidate it.
Keep a no-change or smaller-change option where it is realistic. It gives the team a meaningful reference point rather than forcing a new initiative merely because a list has been requested.
Preserve the decision boundary
The assistant can help organise trade-offs. The authorised owner decides what to investigate or implement. A ranking is not permission to change systems or make commitments.
Save the shortlisted approaches and unresolved questions in the normal planning location. Do not keep polishing the ranking when one missing fact is the real obstacle to deciding. Useful option generation broadens the thinking, then helps the team narrow it with evidence.
For related practical guidance, see understanding the underlying process problem and keeping recommendations separate from authority.
Hili Consulting can help generate genuinely different options before comparing trade-offs. Arrange a free 30-minute consultation and bring one example of the work you want to improve.
