The Use Case Library is a hypothesis generator, not a catalogue to sell from blindly.
Start with the client workflow and evidence. Then use these patterns to ask better questions.
Use the same structure for every opportunity
Problem → Current process → Evidence to collect → Intervention options → Required data → Expected benefit → Complexity → Risks → Human oversight → KPI
Before recommending anything
Ask:
- Should this task exist at all?
- Can the process be simplified first?
- Would rules-based automation solve it?
- Is integration the actual constraint?
- Does probabilistic AI add enough value to justify its risks?
- Where should human approval remain?
High-value patterns to look for
- repeated manual data transfer;
- high-volume drafting and summarisation;
- document classification or extraction;
- knowledge search;
- triage and routing;
- reconciliation and exception handling;
- repetitive decision support;
- handoff delays;
- rework caused by incomplete inputs;
- forecasting or reporting work with large manual preparation steps.
A technically feasible use case can still be commercially weak. Prioritise evidence, value, readiness and implementation path.
