93% of the AI Budget Goes to Technology — and the Outcome Is Decided by the Remaining 7%
There is a paradox running through the finance suites of the world's largest corporations: the organizations investing the most in artificial intelligence are, more often than not, the ones getting the least out of it. Not because of technological failure. The technology works. The problem lies on the other side of the equation — the side that nobody budgeted for with sufficient seriousness.
At the Emerging CFO forum organized by Fortune in partnership with Workday, a group of chief financial officers from Fortune 500 companies, alongside Casey Caram, director and human capital practice leader at Deloitte, placed on the table a figure that deserves sustained attention: organizations allocate, on average, 93% of their investment in artificial intelligence projects to data, technology, and infrastructure, and a mere 7% to enabling people to use those tools effectively. This is not decorative data. It is a diagnosis of investment architecture with direct consequences on return.
What these executives described is not a problem of technological adoption in the technical sense of the term. It is a problem of coherence between what an organization declares as a priority and what its spending decisions reveal it to actually be.
When Spending Builds the Illusion of Transformation
There is an understandable — though flawed — logic behind the 93/7 pattern. Buying technology is visible, quantifiable, and produces a narrative of progress that satisfies both boards of directors and external analysts. An installed artificial intelligence platform, a modernized data architecture, a corporate software license: all are legible signals of movement. Training a team of accountants, redesigning the workflows of a financial planning department, investing in having a professional with twenty years of experience change the way they formulate their analyses — all of that is invisible, slow, and difficult to present in a quarterly dashboard.
Caram articulated it with precision: artificial intelligence capabilities are going to become commoditized. What does not become commoditized is human judgment about which question to ask, which piece of data to contextualize, and which signal to ignore within a growing volume of information. That is the top layer of the competency model he proposed: upon a foundation of traditional financial skills — accounting, forecasting, performance management — sits a layer of data and artificial intelligence literacy, and resting on both of those are what he called the essential human skills: judgment, critical thinking, and the capacity to ask the right questions.
The problem is not that organizations are unaware of that model. The problem is that they approve it in the strategy meeting and then contradict it in the budget allocation.
Marie Myers, Chief Financial Officer of Hewlett Packard Enterprise, described this phenomenon from the inside with a clarity that is rarely heard in public forums. Her team used artificial intelligence to redesign internal operational reviews, reducing manual work and generating visible value at the enterprise level. The result was concrete. And yet, Myers identified the real barrier as something that occurs after implementation, not before: the most experienced professionals — those with the greatest accumulated knowledge — are the ones who most resist changing the way they work. "When we implement new technologies, we spend a lot of time obsessing about the technology, and I think we don't spend enough time thinking about the impact on people," she said. And she added something that functions as an operational principle: "You're not going to generate successful change if you don't bring everyone along with you."
That last sentence is not motivational rhetoric. It is a description of the mechanism by which a nine-figure investment in artificial intelligence infrastructure can produce a marginal return because 15% of the team that was supposed to change their workflow simply did not do so.
The Fracture Between Declared Ambition and Actual Choice
Tim Arndt, Chief Financial Officer of Prologis, offered a complementary perspective that illuminates the strategic dimension of the issue. He described how the CFO role has migrated from administration and reporting toward strategy and business leadership, and how artificial intelligence is accelerating that transition by automating routine tasks and freeing up time for higher-value work. "The expectation now is to be a partner at the executive table," he said, contributing to the construction of strategy rather than simply reporting results.
That evolution is real. But it has a condition that rarely appears in digital transformation presentations: it only occurs if the finance team has the capabilities to occupy that space. A CFO who aspires to be an architect of corporate strategy while leading a team that still operates in manual reporting mode does not have a strategic ambition — they have an organizational contradiction. Artificial intelligence can free up time. It cannot guarantee that that time will be used with sound judgment if nobody invested in developing that judgment in the first place.
Tucker Marshall, Chief Financial Officer of J.M. Smucker, described the process from a more operational perspective. The company is modernizing financial systems, automating workflows, and investing in talent development — from programs for early-career professionals to mid-level hires with experience in data and analytics. And he pointed to something that frequently falls outside transformation plans: communication capability. It is not enough for the finance team to generate better analyses if it cannot translate those analyses into business decisions that are comprehensible to those who make them. Data literacy without the ability to communicate its implications produces strategic silence disguised as technical sophistication.
Noémie Heuland, Chief Financial Officer of Moody's, added a different dimension to the diagnosis. With the growth in the volume of available data, the pressure on finance teams is not merely to generate more metrics but to know which ones matter. She described the risk of what she called KPI overload: the tendency to over-quantify at the expense of strategic clarity. The CFO's role increasingly includes contextualizing data, connecting financial metrics to broader business objectives, and situating them within market dynamics. That is not a technical function. It is a function of judgment — precisely the top layer of the model that Caram proposed, and precisely the one that the average corporate budget is underfinancing.
The Cost of Not Choosing What to Let Go
There is something running through all of these executives' accounts that deserves to be named with precision: no transformation of this kind occurs without someone giving something up. The long-tenured professionals that Myers identified as resistant to change are not resisting out of irrational inertia. They are resisting because for years their expertise was the asset that set them apart, and artificial intelligence threatens the visibility of that asset. Asking them to relinquish it without offering something in return — without redefining the type of value expected of them, without building a bridge toward a different way of working — is asking them to accept a loss with no replacement narrative.
The organizations that are failing at artificial intelligence adoption are not failing at buying technology. They are failing at managing the relinquishment that every transformation demands. And they are failing because that relinquishment does not appear in any line of the budget.
The 93/7 investment model is not a miscalculation. It is an implicit decision that reveals what kind of transformation an organization is actually willing to undertake: the kind that can be measured in installed infrastructure, or the kind that can be measured in real behavioral change. The first produces compelling presentations. The second produces returns.
What these CFOs described at the Fortune forum is not a warning about the future of artificial intelligence. It is a diagnosis of the present state of their own organizations, spoken aloud with unusual honesty. Artificial intelligence is already here. The 93% has already been spent. The question that their companies are now answering — with every training decision that is prioritized or postponed — is whether the other 7% will be enough for anything to actually change.
The organizations that understand that percentage not as a line item of expenditure but as the condition of viability for the return on everything else will hold a concrete advantage over those that continue to treat professional development as a budgetary residual.
