([1])Torsten Slok, chief economist at Apollo Global Management, delivered a stark reality check in an October 24 note. Analysing the “S&P 493”, the S&P 500 index excluding seven big technology firms, he found “no evidence that AI is improving productivity” or boosting profit margins for the rest of corporate America ([2]). In Slok’s words, “the technology has delivered zero financial payoff” to companies outside Big Tech ([3]), despite a surge in spending on AI initiatives over the past year. By contrast, the largest tech players, Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla, have seen revenue per employee and margins rise since the AI boom began ([4]).
Slok’s analysis reinforces what many CFOs of traditional companies have suspected: for now, the bottom-line benefits of AI are largely confined to the tech sector. The finding comes as a sobering reminder that beyond a few AI trailblazers, most industries have yet to translate artificial intelligence into tangible financial gains. For finance chiefs, the message is clear, until an AI project can demonstrate real improvements in revenue, efficiency or profitability, it remains an investment based on faith rather than facts. This underlines the importance of identifying concrete use cases where AI directly drives measurable business value, instead of relying on vague promises of future transformation.
CFOs are increasingly done waiting for elusive returns to materialise. Recent executive surveys suggest that many finance leaders are prepared to pull the plug on AI projects that don’t prove their value on a clear timeline. In one new global poll of CFOs, about half said they would cancel or re-scope any AI initiative that fails to show measurable ROI within the next 12 months, a sign of rising impatience with costly experiments that aren’t paying off. And this hard line comes as AI costs themselves are often running out of control: In April, Uber’s chief technology officer disclosed that 5,000 employees using a generative AI coding assistant had burned through the company’s entire 2026 AI budget by April ([1]). Amazon recently revealed a single “catastrophically expensive” AI project that exceeded its budget by 800% before executives discovered it ([2]).
Finance leaders are responding to such stories by tightening oversight and demanding evidence of impact. In Deloitte’s latest CFO Signals survey, nearly 80% of North American finance chiefs said their teams now have a formal role in approving or monitoring AI spending ([3]). Yet the survey also found that only a minority have advanced, real-time cost controls in place ([4]) ([5]), suggesting many firms still lack the tools to prevent overruns. The overall trend, however, is towards greater accountability: CFOs are setting clearer success criteria for AI investments and standing ready to redirect budgets if those criteria aren’t met. For CEOs, this means any AI project on the go must have specific, trackable benefits, whether it’s reducing cycle times, cutting error rates, or boosting sales, or risk being discontinued.
Even where companies do see gains from AI, they tend to be modest. A new global study by KPMG found that 71% of organisations report their AI investments are meeting or exceeding their expected return on investment ([1]). However, true outperformance is uncommon, only 23% of firms said AI was delivering returns above their initial expectations ([2]). In other words, while most companies using AI are at least getting what they bargained for, only about a quarter are enjoying standout success.
This finding highlights a widening gap between AI leaders and laggards. According to KPMG, broad adoption of AI in finance "is not the same thing" as achieving high-impact performance gains ([3]). Companies that integrate advanced AI, such as “agentic” AI systems that can act autonomously, report significantly stronger improvements in decision-making speed, forecasting accuracy and other metrics than those sticking to basic automation ([4]). For CFOs, the implication is that real ROI on AI comes from focused, well-executed deployments tied to core business outcomes, rather than dabbling in pilot projects. The challenge is to move from ambitious experimentation to scaling the use cases that directly drive revenue growth, cost savings or efficiency, and to measure those gains rigorously.