The AI Payoff Outside Tech Has Not Arrived Yet

Torsten Slok, chief economist at Apollo Global Management, put a blunt number on a feeling that has been building through earnings season. In a note published on October 24, Slok wrote that artificial intelligence had delivered zero financial payoff to corporate America outside the technology sector, according to TechCrunch, which reported on the note the following day. The claim about the AI payoff outside tech landed because it cut against the spending itself. Companies have poured money into AI tools, model access, and infrastructure, yet the broad financial results Slok examined did not move.

Slok built the argument around a split inside the best known stock index. He called the wider group the S&P 493, meaning the S&P 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla. For that wider group, revenue per worker stayed flat, and profit margins stayed flat as well. For the seven largest technology companies, revenue per worker rose and profit margins rose. The AI payoff outside tech, in Slok's reading, had not arrived in the aggregate numbers, even though the largest technology companies were pulling away on the same measures.

The revenue figures behind that split were stark in the reporting on the note. Dealroom reported that revenue per employee for the seven largest technology companies rose from about five hundred fifty thousand dollars in 2022 to nearly eight hundred thousand dollars in 2026, while the wider group stayed near five hundred fifty thousand dollars since 2019. Profit margins told a similar story in that account, climbing for the largest technology companies since the third quarter of 2023 and staying flat for the rest of the index. Those two lines, one rising and one flat, are the core of the AI payoff outside tech debate, because they suggest that gains from AI have stayed concentrated at the top rather than spreading across the economy.

Why Slok doubts a quick catch-up

Slok did not treat that gap in the AI payoff outside tech as a simple lag. TechCrunch reported that he doubted mainstream companies would easily turn cheaper intelligence into higher revenue growth and higher profit margins. He wrote that AI could work well as a calculator, while business intelligence depended on judgment, accountability, trust, culture, unwritten rules, ethical trade-offs, and fairness. Those factors shape reputational risk, and reputational risk shapes revenue and profit margins, according to Slok. In that view, a model that answers quickly does not automatically make a company more profitable, because the decisions that move money still rest on human judgment about customers, employees, and the public.

Slok also pointed to how concentrated AI spending remained. Reporting on the note described Ramp spending data in which a small share of firms accounted for almost all model serving and inference spending, with most of that spending coming from venture-backed companies. Slok wrote that mainstream corporate America was starting to treat token spending as a cost that weighed on margins. That shift matters for the AI payoff outside tech because it changes how finance teams budget for AI. A tool that looked like an investment when every competitor was buying in can start to look like an ordinary expense when the return does not appear in the quarterly results.

The concentration argument also helps explain why workplace stories about AI and company results can both be true at once. A support agent who resolves tickets faster, a marketer who drafts copy in minutes, or an analyst who summarizes a long report before lunch may all be more productive in a real sense. Those gains do not always become wider profit margins, because a company can pass savings to customers through lower prices, spend the savings on implementation and training, or lose them to rising costs elsewhere. Earlier coverage of the gap between perceived productivity and measured results made the same distinction, as readers can see in earlier reporting on the AI productivity perception gap. The Apollo note adds a financial version of that point, measured in margins and revenue per worker rather than in tasks completed.

What the counterevidence shows

The Apollo note on the AI payoff outside tech is a reading of aggregate trends, and other surveys point to a more mixed picture at the level of individual firms. Dealroom reported that a McKinsey survey found more than a third of respondents attributed at least some earnings effect to AI, while four in five said AI improved individual productivity. Deloitte found a similar split in that account, with about two thirds reporting efficiency gains and only about one in five saying AI had lifted revenue. Those results do not contradict Slok directly. They suggest that the AI payoff outside tech may be real inside particular teams and workflows while remaining too small, too uneven, or too recent to move the totals for hundreds of companies at once.

Timing is part of that explanation. General purpose technologies have often taken years to show up in economy-wide productivity statistics, because companies need time to reorganize work around them, retrain staff, and rebuild processes that were designed for older tools. Slok himself framed the question as forward looking, asking whether mainstream firms would convert cheaper intelligence into stronger growth, rather than declaring that AI could never help them. The honest version of the AI payoff outside tech question, then, is not whether any company has benefited, but whether benefits will broaden enough to lift margins across the wider index, and how long that broadening might take.

For workers, the stakes in the AI payoff outside tech question are practical. If AI gains stay concentrated, hiring, pay, and job security may keep diverging between the largest technology companies and the rest of the economy, even as AI tools become standard in offices everywhere. Earlier reporting on how AI speed can strain a workforce, in earlier reporting on workforce fragility, described that strain in human terms, with faster output demands arriving before support and training caught up. Slok's financial lens reaches a compatible conclusion from the other side of the ledger, according to TechCrunch, which reported his view that the technology had not yet changed the bottom line for most large companies. Readers weighing whether to trust a productivity score or a vendor promise can also see earlier reporting on an AI job survival score for a different way of sizing up the same pressure. The Apollo note on the AI payoff outside tech does not settle the argument about AI and work, but it gives skeptics a clear set of numbers to argue from, and it gives believers a clear test to meet: show the AI payoff outside tech in margins and revenue per worker, not only in demos and daily tasks. That test, according to the reporting on the note, is one the wider economy has not passed yet, as TechCrunch reported and Dealroom reported.