Apollo's chief economist finds zero AI payoff outside Big Tech. Make each deployment move revenue per worker or margin, or expect your token bill to be cut

On October 24, Apollo chief economist Torsten Slok published a note titled No Signs of AI Payoff Outside the Tech Sector(opens in a new tab). His charts show revenue per worker and profit margins rising at the seven largest tech companies and flat across the other 493 companies in the S&P 500. He sees no sign yet of AI lifting productivity growth, and he writes that mainstream companies are starting to treat token spending as a cost that weighs on margins, not a source of growth.

The evidence has limits, and they matter. Its two spending charts rest on data from Ramp, whose card and expense records cover Ramp's own business customers, not the whole economy, and the note gives no sample size for them. One shows the top 10% of firms accounting for 99.5% of model serving and inference spend. Aggregate data also lags. A gain from one claims or support deployment takes quarters to show up across 493 companies.

Even so, the reading for an enterprise AI budget owner is plain. Slok is not arguing that AI fails. He is arguing that its payoff is not yet visible in the two numbers a CFO and an investor watch: revenue per employee and margin. If your AI program reports hours saved, adoption rates and pilot counts but cannot point to either line, it sits in the same place as the S&P 493, and the token bill will be read as a cost.

The answer is to decide, before the next budget cycle, which line each funded deployment is meant to move, and to measure it at the business unit where the change can be seen. A support deployment should show cost per resolved case falling. A sales deployment should show revenue per rep rising. A deployment that cannot name its line should be paused, not expanded. That turns an argument about the whole economy into a short list of projects that either show up in the P&L or get cut.