← Resources

The CFO guide to AI investment decisions

Quick answer: CFOs should manage AI as a portfolio of stage-gated investments. Fund small experiments with a fixed budget and end date, require a baseline and an outcome metric before any program scales, review every program quarterly against fully loaded cost and measured return, and make an explicit decision each time: scale, change or stop.

Key takeaways

  • AI spending is forecast to reach about $2.75 trillion worldwide in 2026 (Gartner, November 2026). Budgets are growing faster than the evidence behind them.
  • McKinsey's AI high performers, about 6% of respondents, attribute 5% or more of EBIT to AI (McKinsey, State of AI 2026). Return is possible, but concentrated.
  • The most useful thing a CFO can add is not more AI spend. It is a decision process that moves money from programs that do not work to ones that do.

What makes AI investments different from other technology spend?

Three things. Costs are usage-based and can grow quickly without a new purchase decision. Benefits are spread across many teams and often show up as capacity rather than cash. And the technology changes so fast that a business case can be outdated within two quarters. Those traits call for shorter review cycles and clearer stop rules than a typical software purchase.

How should a CFO structure the AI portfolio?

Sort every AI program into one of three stages, each with its own rules.

Stage Purpose Funding rule Evidence required to move on
Experiment Test whether AI can improve a specific outcome Fixed budget, fixed end date (often 60 to 90 days) A named outcome metric and an early signal against a baseline
Pilot Prove the outcome change in a real workflow Budget tied to a defined group or volume Measured change in the outcome vs a control or baseline, and a fully loaded cost estimate at scale
Scale Capture the return across the business Annual budget, reviewed quarterly Payback on plan, realized value in the P&L, guardrail metrics stable

What questions should a CFO ask before funding an AI program?

  1. Which business outcome will change, and how is it measured today?
  2. What is the baseline, and who owns it?
  3. What is the fully loaded cost at scale, including people and integration?
  4. How will we separate the AI effect from everything else changing?
  5. When will we know if it is working, and what happens if it is not?
  6. If it works, will the value show up as lower cost, avoided cost or more revenue, and in which budget line?

If a sponsor cannot answer questions 1, 2 and 5, the program is an experiment, not a pilot.

How should AI programs be reviewed?

Run a quarterly AI investment review, timed with the financial close. For each program, show fully loaded cost to date, measured outcome change, realized versus capacity value and payback against plan. Then record one decision:

  • Scale: outcome is proven and payback is on plan. Increase budget.
  • Change: outcome is moving but below plan, or cost is above plan. Adjust scope, vendor or design, and set a new review date.
  • Stop: no measurable outcome after the agreed period. End the program and redeploy the budget.

A healthy portfolio stops some programs every year. If nothing is ever stopped, the review is not working.

How should a CFO handle AI budgets for next year?

Budget AI by program, not by vendor. Hold back a share of the total AI budget, for example 15% to 25%, as an unallocated pool that funds the best-performing programs mid-year. That turns the quarterly review into a real reallocation decision rather than a reporting exercise.

What should the CFO report to the board?

A one-page view of total AI spend, spend by stage, the top programs by return, the programs stopped or changed and why, and the next quarter's decisions. Consistency matters more than detail. The same metrics each quarter build trust.

Frequently asked questions

How should a CFO evaluate AI investments?

Treat each AI program as a stage-gated investment with a baseline, a fully loaded cost, one outcome metric and a review date, then decide every quarter whether to scale, change or stop it.

What is a reasonable payback period for an AI program?

It depends on the program and your cost of capital. Many finance teams set a hurdle of 12 to 24 months for operational AI programs, with longer horizons for strategic ones. Set the hurdle before funding.

Who should sit on an AI investment committee?

The CFO or a delegate as chair, the CIO or CTO, a FinOps or IT finance lead, and the business owners of the largest programs.

Should AI budgets sit in IT or in the business?

Either can work. What matters is that each program has one accountable business owner and that finance sees the full cost across all budgets.

Sources

Get updates