A practical method for AI opportunity discovery

Discovery is a decision process

The output of AI discovery is not a catalogue of possible use cases. It is a justified decision about where to invest attention next, what evidence is missing and who must own delivery. A useful discovery reduces ambiguity rather than decorating it.

1. Frame the operational problem

Begin with work: a costly decision, a slow workflow, inconsistent service, constrained expert capacity or an unmet customer need. Record the current baseline, affected users and the reason the issue matters now. Do not assume AI is the answer.

2. Map the decision and information flow

Identify inputs, outputs, hand-offs, exceptions and accountability. This exposes whether the real constraint is data quality, policy, integration, incentives or process design. It also shows where human judgement must remain visible.

3. Test feasibility early

Check whether representative data exists, whether its use is permitted, how performance could be evaluated and where the capability would connect to existing systems. A compelling idea with no evaluation method is not ready for prioritisation.

4. Score value and delivery conditions separately

Potential value should be considered alongside feasibility, risk, adoption effort and strategic fit. Keep those dimensions separate: a high-value idea with poor delivery conditions may justify foundational work, while a modest idea with strong conditions may be the better first release.

5. Design the smallest useful test

State the assumption, the evidence that would support or reject it, the people involved and the time box. A test might be a data review, workflow simulation, controlled prototype or service experiment. The purpose is learning, not a polished demonstration.

The decision pack

A strong discovery finishes with a short opportunity portfolio, explicit assumptions, risk and data findings, recommended experiments, named owners and a 30–90 day action plan. It should also identify ideas that were deferred or rejected and explain why.

Common failure mode

Teams often rank ideas by enthusiasm alone. This rewards novelty and senior sponsorship rather than readiness. The correction is simple: require every opportunity to show a measurable outcome, an accountable owner and an achievable route to evidence.