Method · Aria Zuniga

AI Use-Case Scorecard

Most AI ideas should not be built. This is the rubric I use to decide which ones are worth a team's time, before any money is spent. Score each dimension from 1 (weak) to 5 (strong). The weighting reflects what usually breaks AI projects, and it can be tuned per organization.

Business value & impact

weight 22

How much does this move a real metric? Revenue, cost, time, risk reduction. 5 = clear, measurable, board-level.

Technical feasibility

weight 16

Can current AI actually do this reliably? 5 = proven pattern; 1 = research project.

Data readiness & quality

weight 14

Is the data available, clean, and permitted for use? 5 = ready; 1 = scattered or missing.

Cost efficiency

weight 12

Build plus run cost against the value. 5 = cheap to build and operate; 1 = heavy on both.

Risk & compliance posture

weight 16

GDPR, IT security, reputational exposure. 5 = low risk, well controlled, human-in-loop; 1 = unresolved.

Adoption readiness

weight 10

Will people actually use it? Change effort and trust. 5 = easy fit into existing workflow; 1 = heavy lift.

Time to value

weight 10

How fast can it show results? 5 = weeks; 1 = many quarters.

50/100

Pilot / scope down

Worth a small, time-boxed pilot. Tighten the weakest dimensions first, set a clear go / no-go checkpoint, and keep the blast radius small.

How to read it. The score is a conversation-starter, not a verdict. A low dimension that's fixable (data readiness, adoption) means "scope it down and pilot," while a low dimension that isn't (value, risk) usually means "don't." The point of the tool is to make the trade-offs explicit and force the conversation early, with the people who own the budget.