An AI initiative should be evaluated like any other operational investment: what is the current cost of the problem, what would a solution realistically change, and what would it cost to build and maintain?
Establish the baseline
Document how long the work takes today, how often it happens, who performs it, and the cost of errors or delays. Without a baseline, any projected improvement is guesswork.
Model a range, not a promise
Use conservative, expected, and optimistic assumptions. Present potential annual economic opportunity, estimated investment, potential net benefit, and potential payback period — and label them as estimates.
Include the full investment
Integration work, data preparation, testing, employee enablement, human oversight, and ongoing maintenance all belong in the investment estimate. Leaving them out is the most common way ROI gets overstated.