AI Use Case Business Case

1. Use case summary

One table a reader can take in at a glance. Name a person, not a team, as owner.

2. Problem and intended outcome

What is being solved, for whom, and what should change in the workflow if it works.

• What problem does the workflow have today, and who has it?
• What operational change should result: fewer hours, shorter cycle time, fewer errors, more volume handled?
• What is the primary outcome measure, and what quality floor must the output meet to count?

3. Baseline

The workflow as it runs now, and how each figure was measured. If there is no baseline, say how one will be built before money is committed.

4. Value hypothesis by layer

What this use case will be able to show at each of the five layers of AI value management. The layers are cumulative: a claim at layer 3 needs evidence at layers 1 and 2 first.

5. Full cost

Every cost line, in the seven layers of the AI TCO Framework, not only the model invoice. Include verification and review time under people.

6. Attribution plan

How this spend will be counted as outcome-linked in attribution coverage, and what evidence will stand behind the value claim.

• Which spend lines carry this use case's tag or cost centre, so the spend can be found again?
• Which attribution test applies: a baseline and an outcome measure, or decision-grade evidence?
• Evidence level today and the level targeted, on the E0 to E4 scale. No monetised claim below E2.
• Benefit type: realised cost reduction, revenue or margin, absorbed capacity, released capacity, avoided cost, or risk loss avoided.
• Where the benefit will be booked (a budget line or an operational record), and who in finance accepts it.

7. Risk and controls

How much human attention the output needs, and how hard a mistake would be to undo.

8. Exit criteria and review

Written before the money is spent: the conditions under which the use case stops or is redesigned, who can invoke them and when the first check is.

• Stop if: ...
• Redesign if: ...
• Who can invoke the exit criteria, and the date of the first check.

9. Decision requested

What you are asking for, the alternative considered, and what must be true at the next gate.

• Request: fund the pilot / fund with conditions / do not fund.
• Alternative considered, including doing nothing or a non-AI change to the workflow.
• What must be true to pass the next gate, and the date.