Apollo Global Management’s chief economist Torsten Slok has poured cold water on the idea that the current AI boom is lifting all boats. In a note published 24 October, Slok argued that artificial intelligence has “delivered zero financial payoff to corporate America outside of tech,” pointing out that only the largest "Magnificent Seven" tech companies are seeing rising revenue per employee and widening profit margins from AI adoption ([1]). For the remaining 493 companies in the S&P 500, those productivity and margin metrics are flat ([2]).
This dearth of tangible ROI is already influencing spending behaviour. “With no impact on the bottom line, mainstream corporate America is pulling back and starting to view token spending not as a source of growth but as a cost centre that weighs on margins,” Slok writes ([3]). His warning adds weight to what many finance chiefs have been saying in recent months: boards and CFOs are no longer writing blank cheques for AI initiatives without clear evidence of returns. Instead, they are scrutinising AI budgets and shelving projects that fail to demonstrate measurable value.
CFOs got another reality check on 26 October when advanced AI developer Anthropic made public the massive expenses behind its technology. The OpenAI rival’s newly filed IPO prospectus showed first-half 2026 revenue of $16.3 billion, but a net loss of $80.7 billion, driven largely by the accounting impact of convertible debt and other funding tied to its colossal compute needs ([1]). The scale of these losses, which dwarf the startup’s sales, underscores how expensive it is to build and run cutting-edge large language models.
For finance leaders, Anthropic’s finances illustrate why focusing on costly AI moonshots can be perilous if returns are years away. As Booking Holdings CFO Ewout Steenbergen noted at a summit this month, even the biggest AI "hyperscalers" are essentially guessing at their future ROI ([2]). His advice to user companies: capture value by applying AI to redesign key processes end-to-end, while keeping a tight grip on usage costs ([3]). In other words, let others invest billions in AI development, and leverage their models selectively, with strict cost controls, to ensure any efficiency gains actually translate into financial returns.
Despite muted results at the macro level, there are signs that carefully targeted AI deployments can pay off when paired with disciplined management. In its third-quarter results announced last week, Munich-based Atoss Software reported that AI-driven productivity improvements, mainly in software development, helped lift its operating profit by 23% year-on-year to €21 million, roughly 10% above analyst expectations ([1]) ([2]). On this basis, Atoss raised its full-year operating margin goal to about 37%, up from a previous 34%, bucking the trend of tech investments eroding profitability.
Such examples highlight how the return on AI is most evident in specific workflows where one executive owns a clearly defined metric for success. Earlier this year, the CFO of a large insurance group cited AI-driven automation in customer service for reducing call volumes by 20%, a change credited with delivering over $80 million in annual run-rate savings ([3]). The common thread: these companies targeted narrow areas with high potential for efficiency gains, established a baseline, and tracked improvements rigorously. Leaders looking for AI payoffs would do well to emulate this focused approach, and to continually monitor costs to ensure that any productivity gains truly boost the bottom line.