When AI Spend Stops Behaving Like Software
In April, Uber's Chief Technology Officer told The Information that the company had exhausted its entire 2026 budget for AI coding tools — four months into the year. The tool was not misused. Engineers were doing exactly what leadership had asked: adopting Claude Code as fast as possible, tracked on an internal leaderboard. A few weeks later, Uber's president and COO, Andrew Macdonald, admitted something more uncomfortable on a podcast interview: that the rising spend still could not be clearly tied to more features shipped for customers. "That link is not there yet," he said.
This is not really a story about Uber, or even about Claude. It is a preview of a problem finance teams are only starting to name.
A new kind of cost line
AI spend does not behave like the software budgets finance is used to managing. A SaaS seat is a fixed cost: pick a tier, multiply by headcount, done. AI tooling is priced closer to a utility — billed on usage, on runtime, on how much a model actually gets asked to do. And falling per-token prices do not translate into falling bills, because the tools people build on top of AI use more tokens per task over time, not fewer. The result is a cost line that moves on its own, month to month, without anyone consciously deciding to spend more.
Arriving at the same time as a skills gap
This would be a manageable problem for a finance team with spare capacity. Most do not have it. Robert Half's 2026 Demand for Skilled Talent research found that FP&A is now the single largest skills-shortage area finance leaders report, and that skills shortages have caused project delays for three in four finance leaders over the past year. So the same teams now expected to track and explain a genuinely new, fast-moving cost category are already short-staffed on the analytical capacity that would let them do it well.
The economics are murkier than either side of the debate admits
It would be easy to conclude from all this that AI is either a hiring driver or a cost trap, depending on which vendor's newsletter you read. The honest picture sits in between. Ramp's research, tracking AI spending against workforce data across more than 21,000 US firms, found that companies investing heavily in AI grew headcount over the following two years — but that gain applied only to high-intensity adopters; lighter adopters saw no measurable change. A 2026 survey of 102 investor-backed CFOs found 87% were still hiring finance staff even as AI use scaled. At the same time, Goldman Sachs' own cost analysis this year found that for contact-center work, an AI agent actually costs slightly more per day than a human employee once real usage is accounted for — not less. And Klarna's well-known 2024 decision to replace roughly 700 customer service agents with AI was significantly walked back within about a year, after its CEO acknowledged that cutting too hard on cost had come at the expense of quality. None of this means AI does not pay off. It means the payoff is not automatic, uniform, or especially predictable — which is precisely the kind of judgment call that belongs to finance, not to a vendor's sales deck.
What this actually calls for
Most advice on this topic stops at "govern your AI spend" — a dashboard, a spending cap, a quarterly usage review. That is necessary, but it is not sufficient. A dashboard does not interpret variance. It does not negotiate an enterprise renewal, or explain a sudden run-rate increase to a board. People do that. And the finance teams that will handle this well are not necessarily the ones moving fastest on AI adoption — they are the ones with enough experienced capacity to understand what is actually happening in their AI spend before it becomes a surprise on someone else's desk.
Views expressed are the author's own.
________________________________________
Frédéric Vanduynslaeger is a Fractional CFO and Strategic Advisor to growth-stage financial services companies, with 25+ years of finance and regulatory transformation experience across Crédit Agricole Assurances, BNP Paribas Cardif, HSBC, and Moody's Analytics. His background bridges institutional-grade financial discipline with the operational realities of scale-ups navigating AI adoption.
Focal Point Business Coaching of Brooklyn provides fractional CFO, strategic advisory, and executive coaching services to growth-stage companies in financial services, insurtech, and climate tech — helping founders build investor-grade financial infrastructure without the cost of a full-time hire.