Use Your AI Budget to Fund Your Data Foundation

Companies fund the AI and skip the data underneath it

We’ve all heard the term “burning platform.” Show leadership the fire and the money moves.

Right now, AI is the burning platform. Every leadership team wants use cases in production this year, and every one of those use cases demands accurate data. For those of us who have spent our careers in data, this is a familiar fight with a new face.

We barely have to make the case anymore. We get into a room and the CIO, the analytics lead, and the engineer who maintains the pipelines have already beaten us to it. Fragmented sources. Four competing definitions of a customer. A copilot rollout sitting on top of all of it. Everyone in the room can see the smoke.

Then the budget comes through. It funds the enterprise AI licenses and a pilot with a launch date, while the data underneath them gets nothing.

A data platform’s value comes from how applications use it

Data teams have struggled to make the business case for their platforms for decades. The applications that depend on those platforms produce measurable results, like accurate forecasting or a pricing model that improves margins. Those applications get the credit. The platform rarely moves revenue or cost by itself.

So when the business asks what the platform is worth, the honest answer is “we don’t know.” There’s no way to run last year twice, once with the mess and once without, and compare the two. The benefit gets described instead of counted. Then the platform competes for funding against an application that cuts overhead, with savings somebody modeled to three decimal places.

AI raised the stakes. For years the main consumer of enterprise data was a person, and people compensate. The analyst knows which customer table is wrong and fixes the number before it reaches the deck. But an AI model can process enormous amounts of data without knowing which records are trustworthy. The data platform that once helped analysts reconcile conflicting information now has to provide that consistency before AI ever encounters the data.

AI changes the funding conversation for data teams

AI made the invoice bigger and put more people in the room. This year, one of them is the CFO.

In late 2025, Forrester predicted CEOs would lean on their CFOs to approve AI investments based on ROI, and a quarter of planned AI spend would slip into 2027. By September 2026, KPMG found 74 percent of large US companies running a cost review before approving AI, up from 61 percent a quarter earlier.

That scrutiny gives data teams the opportunity to make the case for work that’s historically been hard to fund. When finance evaluates the cost of an AI use case, include the data preparation required for success in the estimate, from matching customer records to reconciling conflicting definitions. For years, we’ve struggled to find that work a place in a project budget. Now it can be tied to a specific AI investment.

The ROI calculation starts before the dashboard

The most useful number in KPMG’s mid-year global survey is the one not getting quoted. Organizations with full visibility into what AI costs them to operate are five times more likely to report established ROI. Fifteen percent versus three percent.

Most people read that as an argument for a cost dashboard. By September 2026, 70 percent of large US companies were running AI monitoring dashboards.

Now, look at what has to be true before a dashboard can tell you anything about ROI. To attribute operating cost, you need a use case with a boundary, an owner, and a workload you can point to. Somebody scoped a real thing and put a name on it. The dashboard is the fingerprint that discipline leaves behind.

That’s what most requests to make data AI ready are missing. Meanwhile, you’re already paying for the mess. It shows up as the Thursday night reconciliation, the analyst whose real job is checking whether the number is right, the third version of the customer list, and the second pilot rebuilt because it couldn’t reuse anything from the first. That analyst is the filter AI doesn’t have. Those costs rarely showed up in a budget, making it difficult to justify the investment needed to eliminate them.

So quit asking for the foundation

Ask for the outcome and put the foundational work it requires in the estimate.

• Start with the business metric you want to improve: days sales outstanding, forecast variance, cost to serve.
• Choose a use case with a clear line to that metric.
• Work with finance to establish the baseline and validate the expected return, then price the cost and return in the same document before anyone architects anything.
• Build the slice of foundation that use case needs, sized to the outcome.
• Ship it, measure it against what you promised, and fund the next one with the receipt.

The payoff grows as you build more use cases on the same foundation. The definitions, the resolved entities, and the quality gates from the first use case are still standing when you start the second. The second use case costs less. The third costs less again. On a Gantt chart, it reads as a string of small projects. In practice, it’s one foundation, built a use case at a time, with each piece paid for by the team that needed it.

Your next AI meeting should skip the readiness question

Everyone in the room already knows the data isn’t ready. Another discussion about readiness won’t get the necessary work funded.

Ask this instead: Which business number are we moving, what does moving it cost, what does it need underneath it, and who in finance has signed off on the baseline?

If you can’t answer that in dollars, you’re not ready to ask for the money yet. Go do the arithmetic. That part takes a week.

The AI licenses will renew either way. Stop waiting for the fire to reach flashover. Pick the use case somebody needs this year and put the data work in its estimate.