Track 10 · Discovery & handoverLesson 03 of 12
Picking the wedge use case
Qualify and rank use cases for value in weeks, not quarters.
- Lesson library · reference
- 35 min read
- 60 min hands-on
- Pro
What you'll be able to do
01- You can qualify a candidate AI use case against value, feasibility, data readiness, risk tier, time-to-value, sponsor strength and repeatability, and explain each score with evidence.
- You can apply hard disqualifiers before scoring, and explain why an additive score must not be allowed to compensate for them.
- You can choose between RICE, ICE, a value/effort 2×2 and weighted scoring for a given situation, and name the bias each one introduces.
- You can test whether a ranking is stable under reasonable changes to the weights, and say when a "winner" is really a tie.
- You can build a portfolio of one quick win and one strategic bet, and park the rest with a written reason and a revisit trigger.
- You can say no to a sponsor's favourite use case without losing the sponsor.
Before you start
- Finding the real requirement: each candidate should already have a problem statement and success measure.
- Shadowing the work as it's really done: the process evidence you use to size value.
- Agent or workflow? Choosing the autonomy level: the check that decides whether a candidate needs an agent at all.
- Approval gates and bounded autonomy: needed to judge the "human where needed" disqualifier.
The short version
Discovery always produces more candidate use cases than you can build. Qualification is how you pick one wedge that can prove value in weeks: first remove the candidates that cannot work as framed (hard disqualifiers), then score the rest on a small set of weighted criteria, then check whether the ranking survives reasonable disagreement about the weights. You pair the winner, a quick win, with one strategic bet that the quick win helps make possible, and you park everything else with a reason and a condition for coming back to it. The score is a structured argument, not an oracle.
- Disqualify before you score. No ground truth, no tolerance for error with no human to catch it, no data access, or unbounded scope cannot be outweighed by high value.
- Every scoring framework has a bias. RICE favours reach, value/effort favours whatever was estimated optimistically, weighted scoring favours whoever chose the weights.
- Test rank stability. If two candidates swap places when a weight moves by a few points, you have a tie. Decide it on strategy, openly.
- Check "agent vs workflow vs no AI" for every candidate. Some of the best items on the list need no model at all.
- Parking is a decision, not a deferral. Write the reason and the trigger that would bring it back.
Pro lesson
The rest of this lesson is in Pro.
The mental model, worked example, field note, failure table, exercise and self-check, with the voice tutor.
- Every stage and specialisation, in full
- Every lab and every role project
- Halden weeks 2–10, on your record
- Every drill, live in the playground
- All solutions, plus the server judge
- Lab Assistant and every Teach with AI