How do you justify AI investment to a CFO?

Anchor to one workflow with a known cost, not to "adopting AI". State what that workflow costs today in time, error, or delay; what a scoped system to fix it costs; how fast it ships; and that the result is an asset you own, not a recurring subscription. A concrete, owned, weeks-not-quarters case survives finance in a way a platform pitch does not.

Why do AI business cases get rejected?

Because they sell a platform and a productivity guess. The value is vague ("10% more efficient"), the cost is recurring and open-ended, and the risk looks like the 95 percent of organizations that saw no measurable return. A CFO is not anti-AI; they are anti-unbounded-cost-for-unproven-value.

Should AI be a capital or operating cost?

A forward-deployed engagement produces an owned asset that lives in your accounts, closer to a capital investment than a subscription. You pay to build it once and run it yourself, rather than renting it indefinitely, which changes the whole shape of the business case.

What proof should a CFO require before scaling AI?

A running system with real users, and the measured before-and-after on the one workflow it addresses. Value realized on a live system beats projected ROI on a slide, and it is the honest basis for deciding whether to fund the next workflow.

The AI business case a CFO will sign

The short answer

The AI business case a CFO signs is not "adopt AI." It is one workflow with a known cost, a scoped system that ships in weeks, and an asset you own rather than a subscription you rent. Concrete cost, bounded spend, owned result. That case survives finance; a platform pitch with a productivity guess does not.

A CFO is not the obstacle to AI. They are the person asking the question everyone else skipped: what does this cost, what does it return, and what happens if it does not work. Most AI business cases have no good answer, which is why they stall in finance. The ones that get signed are built differently from the start.

Why most cases die in finance

MIT's NANDA research found 95 percent of organizations investing in generative AI saw no measurable P&L return. A CFO has read that number, or felt it. So a case built on platform adoption and a vague productivity uplift reads as unbounded cost for unproven value, which is the exact profile of the failures. The problem is not finance being cautious. It is a business case that gives caution nothing to hold onto.

Anchor to one costed workflow

Value that a CFO can sign is attached to a workflow, not a platform.

Start where the money already is. Name one workflow and cost it honestly: the hours, the error rate, the delay, the revenue it gates. That number is the case. Against it, a scoped system has a bounded price and a delivery date measured in weeks. This is automation ROI done honestly, raised to the level a CFO signs: not efficiency in the abstract, but this cost, removed, for that price. It is also why the entry point is one team, one problem.

An owned asset, not a subscription

The shape of the spend matters as much as the size. A rented platform is a recurring cost with no terminal value: the meter runs forever and stops the day you stop paying. A forward-deployed engagement produces an asset that lives in your accounts and keeps working after the invoice is paid, because you own it: code, prompts, infrastructure, evals. A CFO understands the difference between buying an asset and renting a dependency, and it usually decides the case.

The number that closes it

The strongest line in the case is not a projection. It is the measured before-and-after on the one workflow, taken from a system that is already running for real users. Fund the first workflow to get that number, then let it decide the next one. That is a business case that compounds on evidence instead of optimism, and it is the enterprise argument we make on the enterprise page.