The Two Percent Fix: What Happens When Finance Owns AI Value

I read a Fortune piece this week covering a survey of 1,006 global senior executives, and one number stopped me: only 2% of organizations give their CFO accountability for AI value.

Two percent.

Meanwhile, 86% of companies are increasing their AI budgets this year. The gap between those two numbers tells you almost everything about why the AI ROI conversation feels so stuck right now. When CFOs do own AI value, 76% of those companies report achieving “a great deal of value” from their investments. That’s not a marginal difference. That’s a completely different category of outcome.

Where the Accountability Actually Sits

The survey, conducted by Thomas H. Davenport and Laks Srinivasan and reported by Fortune’s Sheryl Estrada on March 27, examined how enterprises structure AI accountability across industries.

Most companies follow the same playbook. The chief data officer, chief analytics officer, or chief AI officer gets the AI value mandate. About 38% of organizations assign accountability this way. It makes intuitive sense. These are the people who understand the technology, build the models, and run the data infrastructure.

The problem is that understanding the technology and proving it shows up on the P&L are completely different skills.

That credibility gap shapes the entire measurement challenge. 44% of respondents cited generative AI as the most difficult technology to establish value from. Another 24% said the same about agentic AI. And here’s the part that might reframe the whole conversation: 50% of surveyed organizations said they get the most value from analytical AI. Not generative. Not agentic. Plain analytical AI. Only 9% named generative AI as their top value driver.

So the technology categories getting the biggest budget increases are also the ones companies find hardest to measure and least likely to cite as their primary value source. That disconnect deserves more attention than it’s getting.

A separate report from Info-Tech Research Group, also published March 27, found that over 75% of CIOs expect their organizations to invest in agentic AI by end of 2026. George Khreish, the firm’s Managing Partner, put it directly: “CIOs are no longer being asked whether they are investing in AI, but whether that investment is delivering measurable value for the business.”

The question has changed. The accountability structure hasn’t.

The Training Gap That Makes Everything Harder

Here’s where the survey gets uncomfortable. 58% of the organizations represented haven’t trained their employees in basic AI use.

That’s the majority of companies pouring money into AI while the people meant to use it don’t know how to use what’s been built for them.

And it’s not just frontline teams. 29% of executives acknowledge leadership knowledge gaps around AI. When the people approving your budgets can’t evaluate what they’re buying, accountability becomes performance art.

The Info-Tech report reinforced this from the infrastructure side, identifying data governance as the single largest capability gap across the organizations it studied. The foundation needed to measure AI value (clean data and clear ownership) is the same foundation most organizations are weakest in.

The hidden fear among CEOs I talk to: “Am I about to approve the next headline-making AI failure?” When nobody with financial authority is tracking outcomes, that fear makes sense. But the data here suggests the fix isn’t more caution. It’s better structure. And it starts with two decisions: who measures, and who trains.

What the 76% Have in Common

The companies getting real value from AI share three structural traits. None of them involve model selection or budget size.

First, they put financial discipline at the center of AI governance. Not as a cost-cutting exercise, but as a measurement structure. The CFO doesn’t need to understand transformer architectures. The CFO needs to know whether a workflow change actually moved a number that matters.

Second, they train leadership and employees simultaneously. The 23-point advantage for dual training is substantial. Teams can’t adopt what they don’t understand. Leadership can’t govern what they can’t evaluate. Both gaps close together or neither does.

Third, they aggregate value instead of isolating it. The survey found that 72% of high-performing organizations measure AI value across the organization rather than treating each use case as a standalone experiment. The shift from “did this tool save time?” to “did AI change how we operate?” is where real ROI gets visible.

The bar for outperforming the competition on AI value isn't better technology. It's better ownership.

AI spending is migrating from innovation budgets with loose ROI requirements into operational budgets with the same rigor applied to headcount decisions. That’s not a correction. That’s AI growing up.

Over Coffee

If I had ten minutes with a CEO right now, I’d say this: before you approve next quarter’s AI budget increase, answer one question. Who in your organization has both the authority and the incentive to say “this isn’t working” and be believed?

If the answer is nobody, that’s the starting point. Not a new model. Not a new vendor. An accountability structure that includes someone who reads a balance sheet for a living.

The 2% stat is genuinely good news. It means the gap between companies that get AI value and companies that don’t isn’t about talent, technology, or spend. It’s about a structural choice that most organizations simply haven’t made yet. And structural choices are the kind of problem that gets solved in one good meeting.