How to present AI ROI to a CFO — and leave with the budget approved
A framework of hypothesis, timeline and measurement that takes AI out of the "innovation folder" and puts it at the center of the P&L.
A good CFO isn't averse to AI. They're averse to vagueness. Almost every AI project pitch fails for the same reason: it can't precisely answer three questions. How would it work? In how long would we see it? How would we know it worked?
The hypothesis must fit in one sentence
"Reduce credit analysis time from 48h to 4h, keeping the current default rate." That is a hypothesis. "Apply generative AI to improve customer experience" is not. The CFO doesn't fund the second.
The timeline must have milestones
A 12-month project without milestones is a project of undefined size. Break it into 4 ninety-day milestones with clear advancement criteria. The CFO understands this because it's exactly how they look at the rest of the company.
Measurement must pre-exist
"How we'll know it worked" can't be defined later. It has to be defined upfront — which KPI, which current baseline, which target, which acceptable error margin. If that measurement doesn't exist today, the project's first milestone is building the measurement.