Most AI spend still reaches the CFO as a finished business case, with the scope, vendor and expected benefits already settled. By then, finance can approve or decline, but it has little say in what is being built or why.

That is a missed opportunity. Few functions sit across as much of the organisation as finance. It has access to the information, processes and workflows that an AI investment relies on, which makes it a natural co-designer of the investment as well as its approver. Tariq Munir, author of Reimagine Finance, describes the shift as moving from "a controller of an investment" to "an enabler of innovation."

For the CFO, that means co-leading the digital and transformation agenda with technology and product leaders, rather than acting as the final gatekeeper on a business case someone else has already built.

Run AI as a portfolio

A portfolio view gives that role some structure. Each AI investment is tied to a clear return, and the spend is judged as capital allocation, with finance at the table from the start.

That changes how individual business cases are read. Instead of approving or rejecting each one on its own merits, finance can weigh them against each other and against other uses of the same capital. A project that clears the hurdle in isolation may be the weakest option once it sits next to three others.

It also means the portfolio needs a deliberate mix. Some investments should pay back within a year and help fund the rest. Others are longer bets on capability the business will need, and they should be sized so that a write-off does not set the whole programme back.

Capital allocation is also not a one-off decision. Funding is released in stages as each investment shows results, and money moves away from projects that stall toward the ones that are working. Finance already applies this discipline to other capital. AI spend should be held to the same standard.

Hours saved is not a saving yet

Productivity claims are the easiest AI results to report and the hardest to see in the P&L. Hours saved is a useful early measure, but it only becomes value when it turns into lower cost, avoided cost or more output from the same team. The test is a plain question: where are the hours going?

Take 6,000 hours saved in accounts payable. If the team simply absorbs that time, the saving stays on paper. If it means a growing business does not need to add headcount, or the team uses the time to manage suppliers better and capture early payment terms, the benefit is real. If those gains are reinvested in the transformation, the programme can begin to fund itself.

Start with processes, then choose the tool

We would give CFOs the same advice Munir does: "Don't start with AI. Start with your processes." Map where value sits across current workflows and where it is being lost, then decide what each problem needs. Sometimes that is an AI solution. Often it is a process redesign, a systems fix or a change in who does the work. Starting from the process keeps the portfolio honest, because every investment has to show which problem it solves before it competes for funding.

Before you approve budget for the next AI business case, you should be able to answer:

• Which process does this improve, and is AI the right fix for it?

• What return is it tied to, and is that the best use of this capital?

• Is this a near-term payback or a longer bet, and does the portfolio still have the right balance of both?

• Where will the saved hours show up in the P&L, and who is accountable for that?

• When will you review the funding, and what result releases the next stage?

• What result would make you stop?

Not sure which AI investments to fund first

The Omnia Collective works with financial services and enterprise organisations across Australia as a Snowflake and Databricks partner, building the governed data foundations and business definitions that AI investment depends on. See our data engineering and analytics services or get in touch.

This post is adapted from Episode 20 of Nap Stack, Monica’s podcast on AI, data, and building a business. Listen on Spotify or Apple Podcasts.