Your CFO Is About to Ask What the AI Actually Bought. Most Marketers Will Not Have an Answer

I have sat on both sides of this table. Client side, watching a marketing line item balloon with nobody able to explain it. Agency side, being the one who added the line and hoped no one asked. So when I say the reckoning is coming, I am not guessing. I have watched it build.

Here is what I keep seeing when we score Australian businesses on their marketing health. We have measured the marketing of hundreds of businesses across dozens of industries on six dimensions. Almost everywhere, the weakest dimension is the same one: Data and Tracking. Most businesses cannot connect what they spend to what it returns. Not roughly. Not directionally. At all.

That was true before AI. AI did not create this problem. It bolted a bigger, faster spending engine onto a business that still cannot read its own instruments.

For years, marketing got away with it. Budgets crept up, results stayed fuzzy and finance let it slide because the numbers were small enough to ignore. Marketing was never a thinking thing. It was a dreaming thing. You spent, you hoped, you built a nice deck.

AI killed the excuse that it was small enough to ignore.

In Australia right now, marketers are burning through expensive frontier AI models on tasks that do not warrant them. IBM's 2026 benchmark report on AI in Australian marketing put it perfectly. You are using a jackhammer to open a walnut. Every walnut costs more than it should, the token burn climbs month after month and the CFO is watching a line that goes one direction. Up.

The Gartner 2026 CMO Spend Survey tells the wider story. Marketing budgets sit at 7.8% of company revenue, barely moved from 7.7% the year before and roughly 18% lower than they were four years ago. CMOs are now carving out 15.3% of that flat budget for AI. But only about 30% of them say they are actually ready to scale it.

Read that again. You are spending a bigger slice of a shrinking pie on a tool most of you cannot yet operate at scale.

This does not land evenly. A backyard software developer burning a few hundred dollars a month on AI is a rounding error. A midsize fintech running it across every team is a real line finance will question. An enterprise doing it at scale is a board-level number. Same mistake, three very different problems. The fix is identical at every size. Know what the spend does before you have to defend it.

The reason this ends badly is not the AI. It is that marketing has spent years grading its own homework.

The agency says the campaign worked. The platform dashboard says the campaign worked. The tool you bought reports on itself and, surprise, it worked. Everyone in the chain is incentivised to tell you it worked. None of them are the person who signs the cheque.

I sat in a review not long ago where the head of marketing walked the room through a quarter of AI spend. The deck was beautiful. Slide after slide of outputs. Content produced, variants tested, hours saved. Then the CFO asked one question. Which of these sold anything? The room went quiet. Not because the answer was bad. Because there was no answer. The tools had reported on themselves all quarter and nobody had checked the reports against the bank.

Your CFO does not care that your engagement rate is up. A CFO has three questions, and they are the only three that matter. Does it make volume? Does it make margin? Or does it keep the regulators happy? If your AI spend does not do one of those three things, you have no answer to the obvious follow-up. Why are we paying for it?

Here is the test I would run on every line, AI or not. If you turned it off, would anything change? If you cut it and revenue dropped, now you know something. If you cut it and nothing moved, you just found the money finance is coming for. That is not a threat. That is a gift. It tells you where the waste is before someone else does.

A lot of marketers are not great with numbers. That is well documented and it is not an insult. It causes real anxiety. They are scared to sit down with finance because they do not have the answer, and the longer they avoid the conversation the worse it gets. So they slide blind. They keep the spend running because turning it off feels riskier than leaving it on.

That instinct is exactly backwards.

The businesses that survive the coming squeeze will not be the ones with the cleverest AI stack. Nobody in the boardroom cares how many models you are running. They will be the ones who can look the CFO in the eye and show what every dollar bought.

Here is the part marketers keep getting wrong about that room. Finance is not the enemy. We are both paddling the same way. We both want the business to win. A CFO clawing back AI budget is not attacking marketing. They are doing the job marketing refused to do, which is asking whether the spend earns its place.

First, know your two numbers. Do you know how much money you make? Do you know how much money you spend? Everything works back from there. You do not need a perfect attribution model. You need to stop flying blind.

Second, put your AI spend on trial. List every model, tool and subscription. Next to each one, write what it is meant to make happen in commercial terms. Not "efficiency". Volume, margin or compliance. If you cannot fill in the box, that is your answer, and it is the same answer finance will reach.

Third, rightsize the tool to the job. You do not need a frontier model to write a subject line. Match the cost of the tool to the size of the problem. Stop swinging a jackhammer at walnuts. The token bill is not abstract. It is real money leaving the business every month.

Fourth, fix the instruments before you buy more engines. If Data and Tracking is your weakest dimension, and for most businesses it is, then no amount of AI spend is safe, because you cannot read what it does. Get the tracking right first. Then you can scale when you know, and scale back when you do not.

I do not think the budget squeeze is a bad thing. It is a filter. It separates the marketers who know their numbers from the ones who have been coasting on a friendly dashboard and a hopeful deck.

The ones who know their numbers will walk into the finance meeting and turn it into a case for more. Here is what we spend, here is what it returns, here is what another 2% would do. The ones who do not will get cut, and they will call it unfair, and it will not be.

AI did not break marketing accountability. It just made the bill big enough that someone finally read it. If you can prove what the money did, this is the best year you have had in a decade. If you are guessing, you are getting lucky, and luck is the first thing a CFO cuts.

Know your numbers. The meeting is already in the diary.