AI for CFOs: the budget, the business case and the measures that hold up
The build is the small number. Licences, usage, integration upkeep, review time and training recur every month, and a business case that leaves them out will look better than the result.
A CFO's job with AI is to fund it in stages, count the full cost and insist on a measurable return. Budget every use case with one-time and recurring lines, tie it to a number the business already tracks, release money at stage gates, and report results against the original target. Savings that are not removed from a budget or redeployed to named work are not savings.
NetSuite
QuickBooks
Microsoft 365
Google Workspace
Anthropic
OpenAI
Snowflake
Where the numbers come from
These counts come straight from the lists on this page. Nothing here is a survey result or a client outcome.
How do you test an AI business case?
Ask for the baseline, the mechanism and the owner. A case that cannot state today's number, how AI changes it and who is accountable is a hope, not a case.
- Baseline: what is the number today, and where does it come from?
- Mechanism: what exactly will change in the work, and for whom?
- Full cost: are all three kinds of cost in the table included for three years?
- Realisation: where will the benefit appear in the accounts or the plan?
- Owner: who signs up to the target, and when is it reviewed?
How should the return be counted?
In the same terms as any other investment, and in three separate buckets so soft benefits do not hide a weak case.
- Cash: cost removed from a budget, or revenue that can be traced to the change.
- Capacity: hours freed and redeployed to named work. Count it only when the redeployment is real.
- Risk and quality: fewer errors, faster response, better compliance. Report it, but do not add it to the cash figure.
Productivity gains that are not captured, by a lower budget or by more output from the same team, will not show in the results however real they feel.
Which measures should finance report?
A short set per use case, fixed at funding and reported every quarter until the case is proven or stopped.
- The business result against target.
- Adoption: the share of the intended users who use it weekly.
- Unit cost: total running cost divided by the work done.
- Quality: the rate at which outputs are corrected or rejected by reviewers.
- Payback: months to recover the one-time cost from the cash benefit.
Where does AI help the finance function itself?
In work that is frequent, rule-heavy and already reviewed by a person: matching, coding, drafting and explaining. Keep judgement, approvals and sign-off with people.
- Coding and matching transactions, with exceptions routed to a person.
- Drafting variance commentary from the numbers for a controller to edit.
- Chasing and summarising collections activity.
- Answering staff questions about policy from the policy documents.
Questions executives ask about this
Should AI be capital or operating expense?
That depends on what is being bought and on your accounting framework. Much AI spend is subscription and usage and is expensed. Ask your auditors about any build costs you are considering capitalising; this page is not accounting advice.
How do you budget for usage-based costs?
Estimate volume from the baseline, price a low, expected and high case, and set an alert and a cap with the vendor or in your own tooling. Review it monthly for the first quarter.
What is a reasonable payback period?
There is no universal figure. Set it per use case when it is funded, and compare it with what you would require of any other investment of that size and risk.
How do we stop paying for licences nobody uses?
Buy seats in stages, check usage after thirty and ninety days, and reassign or cancel what is idle. Make a named person responsible for the licence count.
Primary sources to check against
Frameworks, laws and product details change. Confirm the current text with each publisher and with your own counsel. Consultants in a Box is an independent consultancy; this page is not published or endorsed by any organisation named. Names and logos are trademarks of their owners.
Related guides
AI readiness assessment
Most AI programs that stall were never short of tools. They were short of a clear business problem, usable data or an owner. A readiness assessment finds which one before the money is spent.
Read the guideAI adoption strategy
A pilot proves a tool can do something. Adoption is when a team does its work a different way and a number moves. The gap between the two is ownership, data and habit, not technology.
Read the guideBoard oversight of AI
A board does not need to understand how a model works. It needs to know where the company uses AI, what could go wrong, who is accountable and whether the investment is paying back.
Read the guideAll AI for executives guides
The full leadership playbook, from readiness to agents.
See the hubWant to talk this through for your company?
Book a call and we will go through where you are, what to decide first and who should own it. Or start with the free business analysis.
