Written for the CFO.
For a finance leader, AI is a value-creation lever or a cost with no return — and which one it becomes is decided before the spend. Here’s how to qualify AI investment the way you’d qualify any other.
Every CFO is being asked to fund AI, and the business case usually sounds excellent: efficiency, automation, margin improvement, professionalisation, value at exit. Much of that is genuinely achievable. But a finance leader’s job is to discount the pitch and interrogate the assumptions — and AI business cases tend to hide the same two flawed assumptions that turn value creation into a write-off.
The two assumptions that sink AI business cases
“AI will fix the process.” It won’t. AI’s core effect is speed, and speed only creates value when the underlying process is sound. Accelerate a clean, well-owned workflow and you get a real return. Accelerate a messy, undocumented, poorly-owned one and you don’t fix it — you amplify it. More speed on broken foundations doesn’t create value; it manufactures faster mistakes. A business case that assumes AI will tidy up a weak process as it accelerates it is assuming the one thing AI doesn’t do.
“The pilot economics will hold.” They frequently don’t, because pilots flatter. A controlled pilot with clean, curated inputs produces impressive numbers. Production — messy real data, full volume, edge cases, integration friction — tells a different story. Many AI business cases are, in effect, extrapolating a demo. The finance discipline is to ask whether the pilot’s economics survive contact with reality, and to fund on the answer, not the demo.
What has to be true before serious spend
Before signing off, a handful of conditions need to be honestly true — and they’re value-creation conditions, not technical ones:
The data is good enough to produce output you’d actually rely on for a decision or a number.
The target process works, and is owned. AI amplifies a working, owned process; it can’t rescue a broken or orphaned one. If you couldn’t do the job well manually, AI won’t do it for you.
The foundations are sound enough that acceleration compounds value rather than fragility — a live concern in scaling and PE-backed businesses, where operational fragility rises with growth.
Someone owns it daily. AI only delivers durable value when a named person manages it. Unowned AI decays into shelfware and risk.
Where those hold, AI spend compounds into genuine value. Where they don’t, it mostly buys faster mistakes and a tool nobody runs.
The leadership question
The CFO’s question isn’t “which AI should we buy?” It’s: are we funding acceleration of something sound — or paying to reach a broken process’s failure faster? And: is our first spend actually a tool, or is it getting ready to spend well — a policy, a workshop, or a person to own it?
Try this prompt
Interrogate a proposed AI investment:
“Act as a sceptical CFO reviewing an AI business case. Here’s the proposal: [describe the use case, the claimed benefit, and how it was piloted]. Challenge the assumptions: does this accelerate a sound process or a broken one, will the pilot economics survive production, what data and ownership does it depend on, and what has to be true for the return to be real. Tell me what I should require before approving spend.”
What to do next
Qualify AI spend the way you’d qualify any investment: test whether it accelerates something sound, whether the pilot economics are real, and whether ownership exists. Where the answer is “not yet,” the highest-return first spend is often not software but readiness — a decision about who owns AI, and getting the foundations sound. That’s a value-creation move in its own right.
In closing
For a CFO, AI is one of the clearer value-creation levers available — and one of the easier to turn into a write-off by funding speed onto broken foundations. Managed well, it’s value; funded carelessly, it’s cost with a good story attached.
If your finance leadership would value help qualifying AI investment properly — and deciding whether the first move is a tool, a policy or a person — that’s exactly the conversation Savant and Axulu are built for, including access to fractional and interim leaders who can own the programme and make the value real. Particularly relevant for PE-backed and founder-led businesses where value creation is the whole point.