Scoring and Choosing Use Cases
Rank candidate AI use cases by value, feasibility and risk before committing a team to one.
Scoring and Choosing Use Cases
Customers usually arrive with a list of AI ideas, or one favourite. Scoring turns that list into a defensible choice. Each candidate is rated on value (how much a measurable outcome improves), feasibility (data, access, integration, skills) and risk (regulation, harm if wrong, reversibility). The best first project is rarely the most exciting one; it is the one that proves value fastest with the least risk.
Why it matters
Picking the wrong first use case is the most common reason enterprise AI programmes stall. A flashy but infeasible project burns months and credibility; a feasible but trivial one proves nothing. A transparent score also lets you say no to a sponsor's favourite idea with evidence rather than opinion.
How to do it
List candidates with the people who do the work → define each as a job with a measurable outcome → score value, feasibility and risk on a 1–5 scale with written reasons → weight the scores with the sponsor (often value 40%, feasibility 35%, risk 25%) → check the top two against data availability before committing → write down why the others were not chosen.
A transparent scoring model
from dataclasses import dataclass
@dataclass
class UseCase:
name: str
value: int # 1-5: size of measurable improvement
feasibility: int # 1-5: data, access, integration readiness
risk: int # 1-5: 5 = lowest risk (reversible, unregulated)
WEIGHTS = {"value": 0.40, "feasibility": 0.35, "risk": 0.25}
def score(u: UseCase) -> float:
return round(u.value * WEIGHTS["value"] + u.feasibility * WEIGHTS["feasibility"] + u.risk * WEIGHTS["risk"], 2)
candidates = [
UseCase("Auto-approve claims", value=5, feasibility=2, risk=1),
UseCase("Claims document intake", value=4, feasibility=4, risk=4),
UseCase("Adjuster case summaries", value=3, feasibility=5, risk=5),
]
for u in sorted(candidates, key=score, reverse=True):
print(f"{score(u):.2f} {u.name}")In a real engagement
Run scoring as a workshop with the sponsor, a domain expert and someone from IT or security in the room. The conversation that produces each score is worth more than the number. Keep the sheet: when priorities change later, you can show why the original choice was made.
Challenge
Take the three candidates in the code example. Argue for and against the top-ranked one in front of a sceptical sponsor who wants auto-approval. What evidence would change the ranking?
Checklist
Production checklist
0 of 5 checked
Knowledge check
A sponsor's favourite use case scores high on value but low on feasibility. What is the best response?
Complete
You have finished “Scoring and Choosing Use Cases”. Continue to the next chapter or practise in the FDE simulator.
Mark this chapter as complete
Track your progress and unlock the next chapter.