Working template

AI business case: one-page approval worksheet

Fill this with measured operating evidence. Blank space is deliberate: if a claim cannot be traced, it is not ready for approval.

Decision

What funding or operating decision will this document support?

Name the start event, accepted output, owner, eligible volume, and system boundary.

Record four weeks of volume, active time, elapsed time, rework, error severity, and cost per accepted output.

State exactly what AI prepares, recommends, or executes—and what a person still reviews.

Separate hard savings, capacity, revenue, and risk. Never add released hours to cash savings.

Include licenses, integration, data work, evaluation, review, training, monitoring, incidents, and exit.

Label every material claim A: measured here, B: comparable evidence, or C: assumption/vendor claim.

Define population, comparison, sample, owner, dates, and the outcome, quality, risk, and economic measures.

Write the quality, safety, privacy, or cost threshold that pauses the pilot.

Name who decides to scale, modify, or stop—and what evidence they will receive.

Model seat cost and released capacity, then return here to classify what becomes cash, capacity, revenue, or risk reduction.