The AI payoff remains difficult to see for most large US companies outside the biggest technology firms, according to Apollo Global Management chief economist Torsten Slok. In a note published on October 24, Slok said revenue per worker and profit margins were rising among the Magnificent Seven technology companies but remained broadly flat across the other 493 companies in the S&P 500.
Slok's analysis, published in Apollo's Daily Spark, compares the seven largest technology companies with the rest of the S&P 500, which he calls the S&P 493. He also said the available productivity data did not yet show a clear boost from AI. His conclusion is an interpretation of economic indicators, not proof that AI has produced no benefit at any individual company.
Why AI spending has not clearly translated into returns
Companies are investing in AI tools, models and computing capacity, but the financial benefits are harder to identify outside the technology sector. Slok argues that some businesses are beginning to treat AI model usage, often billed through tokens, as an operating cost rather than a direct source of growth. Token usage measures how much text or other data a model processes and generates, and can influence the cost of using AI services.
The gap matters because higher spending alone does not demonstrate that AI is improving productivity or profitability. Businesses need to show that AI helps them increase output, raise revenue, reduce costs or improve margins enough to justify the expense.
The evidence has limits
Apollo's note presents charts comparing revenue per worker, profit margins and productivity indicators, alongside data about AI spending. Those measures provide context for the wider economy, but they do not isolate AI as the cause of every change in company performance. Other factors, including demand, pricing, labor costs and investment cycles, can also affect revenue and margins.
The comparison also groups hundreds of companies from different industries together. Some individual businesses may be getting value from AI even if the broader measures have not moved. The data therefore points to a gap between AI investment and visible results across much of corporate America, rather than proving that every non-tech AI project has failed.
What businesses need to measure next
For companies adopting AI, the practical question is whether a tool improves a defined business outcome. Useful measures can include time saved on a task, error rates, customer response times, cost per completed workflow and revenue generated. Those results should be compared with the cost of subscriptions, model usage, integration and human review.
Saganote has covered other signs of the tension between AI investment and business outcomes, including Meta's plan to build AI agents faster and its later concerns about progress, as well as AI's reported role in job cuts across employers. These stories describe different developments, but both underline why AI spending and adoption figures do not by themselves establish a return on investment.
Apollo's October 24 analysis leaves the central question open: can companies outside the technology sector turn increasingly accessible AI tools into measurable improvements in revenue, productivity and profit margins? The available indicators cited by Slok do not yet show that broad shift.
