The real problem isn't AI — it's the conversation with the board
Many AI projects die before they start. Not because the technology doesn't work, but because the person proposing them doesn't speak the CFO's language.
The CFO isn't rejecting AI. They're rejecting uncertainty. They're rejecting projects that have no numbers, no bounded pilot, and no answer to the basic question: what happens if this doesn't work?
If you walk into the boardroom with a presentation that talks about "digital transformation" and "automation capabilities", the response will be: "Interesting. We'll revisit this next quarter." And next quarter never comes.
What works is different. It's concrete, bounded, and it talks about money.
First: quantify the cost of the current problem
Before talking about AI, talk about the problem that exists today. The CFO needs to understand how much the current situation is costing, not how much could be gained from a new one.
Take a specific process. For example: the monthly financial close reporting cycle. Ask yourself:
- How many hours does the team spend consolidating data from different sources?
- How many errors are detected after the report has already been distributed?
- How many hours of review and correction does each error generate?
- What decisions are delayed because the data isn't available on time?
With those numbers on the table, the cost of the problem stops being abstract. If the finance team spends 40 hours per month on consolidation and correction tasks, and the average hourly cost of the team is 35 euros, the monthly cost of that process is 1,400 euros. Over a year: 16,800 euros. In that one process alone.
That number is the starting point. The CFO can debate it, adjust it, challenge it. But they now have something concrete to work with.
Second: structure the ROI in three layers
A business case that works doesn't promise a single number. It presents three scenarios with explicit hypotheses.
Layer 1 — Direct time savings
How many hours are freed up if the process is partially automated? In the example above, if an AI agent consolidates 70% of data sources without manual intervention, the team goes from 40 hours to 12 hours per month. Estimated savings: 28 hours × 35 euros = 980 euros per month.
Layer 2 — Error reduction and rework
Errors carry a cost that is rarely measured. If each reporting error generates two hours of review and correction, and four errors occur per month, that's eight additional hours. Eliminating 80% of those errors frees up another 6.4 hours per month.
Layer 3 — Value of faster decisions
This is the hardest to quantify, but the most relevant to the board. If the CFO receives the report three days earlier, what decision can they make that they can't make today? There isn't always an exact number, but the question alone reframes the conversation.
Presenting these three layers with ranges — not exact figures — is more honest and more credible. "We estimate savings of between 900 and 1,400 euros per month on this process, with a six-week pilot that confirms or adjusts that figure" is a statement the CFO can evaluate. "We will increase operational efficiency" is not.
Third: propose a bounded pilot, not a large project
The most common mistake is presenting a full transformation project when the board doesn't yet trust the technology or the vendor.
What a skeptical CFO approves is a pilot with three characteristics:
Defined scope. One process, one team, one measurable outcome. Not "operations automation", but "data consolidation agent for the monthly financial close in the finance department".
Bounded investment. The pilot must carry a cost the CFO can approve without escalating to the full board. In mid-size companies, that threshold typically falls between EUR 8k–15k depending on the industry and size.
Explicit success criteria. How do you measure whether the pilot worked? "At least 50% reduction in consolidation time within six weeks" is a verifiable criterion. If it's met, the board has evidence to approve the next phase. If it isn't, the risk was contained.
Fourth: anticipate objections before they're raised
The board will ask three questions. Prepare for them before you enter the room.
"What happens if it doesn't work?" The pilot is designed to answer exactly that. The investment is bounded, the scope is limited, and the success criterion is clear. If it doesn't work, it stops. There is no three-year contract at stake.
"How much time will this require from the team?" In a well-structured pilot, the team spends between four and eight hours in the first two weeks to map the process and validate the data. After that, the agent does the work. The team doesn't stop.
"Can't the IT team handle this internally?" They could. But do they have the available capacity right now to deliver a production-ready result in six weeks? If the answer is no, an external pilot makes sense. If the answer is yes, the pilot can be run with external support while the internal team learns the method.
A concrete example: industrial manufacturing company, 180 employees
An industrial-sector company with operations in Spain was manually consolidating production, logistics, and finance data at every monthly close. The process involved three people over four days. Estimated process cost: between 2,200 and 2,800 euros per month, not counting cross-system errors.
The pilot consisted of an agent that connected the three data sources, ran validations automatically, and generated a draft report in a reviewable format. By week six, the process went from four days to under four hours. The team redirected that freed-up time to analysis, not consolidation.
The business case the board approved didn't promise that result. It promised to validate it. That distinction is what makes a CFO sign.
Conclusion
The CFO doesn't need to believe in AI. They need to see a business case with reasonable numbers, a bounded pilot, and a clear success criterion. If you come prepared with that, the conversation changes.
If you'd like to review this structure applied to a specific process in your company, you can request a free diagnostic. No commitment, no prior call required. Just a short form and a response in under 48 hours.
