How to Pitch Your CFO on an AI Budget

The first wave of AI spending at many organizations was approved on potential. Leadership had read the headlines, a few licenses didn’t cost much, and nobody wanted to be left behind. The next wave is different. CFOs have seen the first invoices and want to know what they bought. If you’re asking for a real AI budget now, you need an AI budget business case that survives the questions a finance leader will ask.

The good news is that CFOs aren’t against AI. They’re against spending they can’t evaluate. Give them a clear cost, a conservative estimate of value, a way to check whether it’s working, and a way to stop if it isn’t, and most will engage seriously. This post covers how to build that case.

The five questions every CFO will ask

1. What will it cost, all in? Not just licenses: build time, running costs, training, and the internal hours.
2. What will we get, and how will we know? A measurable outcome, compared against a baseline.
3. What’s the downside? What could go wrong, and what limits the damage?
4. What happens if we don’t? The cost of waiting, stated honestly.
5. Why this amount, and why now? The reasoning behind the number and the timing.

Build the case around those five, and most of the conversation takes care of itself.

The asset: a one-page AI business case

Seven sections, one page. Anything longer belongs in an appendix.

1. The problem, in business terms

Start with a process, not a technology. “Our service desk spends about this many hours a month on password resets and access requests, and our median resolution time is this long.” Use real numbers from a baseline. If you don’t have them, capture them first; baseline before you build shows how.

2. The proposal

What you’ll do, for whom, and in what phases. Name the use cases from your use-case register and the business owner for each.

3. The total cost

Licenses or subscriptions; build time, internal or external; usage-based running costs, with the estimate method stated; training; and internal time. Show the first year and the ongoing annual cost separately. Usage-based costs need their own line with the assumptions behind them; budgets and alerts for usage-based AI spend covers estimation and controls.

4. The expected value

Use a simple formula: time saved per task, times volume, times loaded cost, times a realization factor for how much freed time actually becomes valuable work. Choose a conservative realization factor and say so. Add any direct cost avoidance and quality improvements you can quantify. Anecdotes don’t survive budget review covers the method in detail.

5. Risks and controls

Name the main risks, such as data exposure, low adoption, or cost overruns, and the control for each. Include the risk of doing nothing: employees using AI tools without guidance, and AI features switching on inside software you already pay for.

6. Measures and decision points

What you’ll measure, when you’ll report it, and the points at which funding continues or stops.

7. The ask

The amount, the period, and what’s released when.

Ask for staged funding

The most effective structure I’ve seen is staged funding. Instead of asking for a full-year program budget, ask for enough to run a pilot, with a defined decision point and the next tranche released only if the pilot meets agreed targets. CFOs value options: the right to continue if it works and to stop if it doesn’t. A staged request also gets approved faster, because the amount at risk is smaller and the decision to continue is based on your own results rather than vendor promises.

Speak the CFO’s language

• Payback period: how many months until cumulative value exceeds cumulative cost.
• Return on investment: annual value minus annual cost, divided by annual cost.
• Operating versus capital spending: most AI costs are subscriptions and usage, so they’re operating expenses; say so.
• A range, not a point: show a conservative, expected, and optimistic case, and base the ask on the conservative one.
• Stated assumptions: every number in the case should trace back to an assumption the CFO can challenge.

What not to say

• “Everyone is doing it.” Competitors’ spending isn’t a business case.
• “It will boost productivity.” Too vague to evaluate. Name the process and the number.
• “It saves thousands of hours.” Hours saved aren’t dollars saved unless you can say where the time went.
• “The vendor says it pays for itself.” Vendor case studies are marketing. Use your own baseline.

Objections to expect, and answers

“Isn’t this included in what we already pay for?” Sometimes partly. Check your current licensing first and say exactly what’s included and what isn’t. It builds credibility.

“What if usage costs spike?” Show the budgets, alerts, and quotas that cap exposure, and the kill switch.

“Can’t we wait a year until it’s cheaper and better?” Some things can wait. But the foundations, such as permissions, policy, and training, pay off whatever tools you choose later, and meanwhile unmanaged AI use is growing.

“How do we know it’s working?” Point to the measures and decision points in section 6.

Presenting it

Plan for ten minutes, not an hour. Open with the problem and the baseline numbers, because that’s what makes the rest credible. Bring the business owner, and let them describe the problem in their own words; a finance leader hears a department head differently from IT. Walk through cost, value, and risk briefly, with the one-page case in front of everyone and the detail in an appendix. Then end with the specific decision you need: the amount, the period, and the date of the first decision point. Meetings that end with a clear ask get clear answers.

After approval

Report on the schedule you promised, including when the news is mixed. The first report back after approval does more for your next budget request than anything in the original case. A CFO who sees a pilot measured honestly, and stopped or adjusted where the evidence said so, will fund the next one faster.

Bring finance in early

The best time to involve finance is before you write the case. Spend an hour agreeing the method: the loaded cost per hour, the realization factor, which costs count, and how results will be reported. When the numbers arrive, the conversation is about the results rather than the method. And if the pilot succeeds, you’ll have a finance partner who helped design the evidence and is ready to defend it.

Where this fits

The business case is one part of a bigger conversation about direction. If leadership is asking the broader question, the IT director’s guide to answering “what’s our AI strategy?” covers how to frame it, and the training line in your budget is argued in the case for a standing AI training budget.

Where does your team actually stand?

A readiness score is useful evidence in a budget conversation: it shows what needs fixing before money is spent. The free AI Readiness Score uses 10 of the 24 assessment questions and gives you a score in a few minutes.

Get your free AI Readiness Score →

The sample report includes budget guidance written for a CFO audience. Flip through all 38 pages.

Related guides

• Baseline Before You Build: Metrics That Make AI Pilots Fundable
• Turn Your Painful-Process List Into an AI Use-Case Register
• Budgets and Alerts for Usage-Based AI Spend
• Anecdotes Don’t Survive Budget Review: Measuring AI Value
• The IT Director’s Guide to Answering “What’s Our AI Strategy?”