Apollo's chief economist took the S&P 500, removed the seven biggest technology companies, and looked at the other 493. Their margins and revenue per worker are flat, while the seven tech giants' are rising. His conclusion was that AI has paid off nothing outside tech. The numbers are solid, but the conclusion goes further than they do.
Flat can hide a busier worker
Revenue per worker is measured in money. Take a claims team at a regional insurer. With AI drafting letters and checking documents, each handler closes more claims a week. If rival insurers do the same and customers compare quotes, the insurer has to cut premiums or promise faster payouts to keep its customers. Revenue per employee ends up where it started, and so does the margin. Policyholders pay less and wait less, and a chart of the company's results shows nothing.
So flat margins fit two different stories. In one, AI does nothing. In the other, it works and the benefit is competed away. The numbers cannot tell them apart. The second story is what happened in 1990s retail, when Walmart's rivals copied its computer-run supply chain and the gain reached shoppers as lower prices.
Competition decides who keeps the saving
When one company cuts its costs, it can cut its price to win customers, and its rivals have to follow. Once all of them have the same tool, everyone's margin is back where it was and the customer holds the saving. Analysts have made this argument about earlier technologies: a tool everyone can buy stops being a way to stand out and becomes a cost of doing business. The same logic has been applied to AI, where competition tends to erode productivity gains and hand them to customers.
The economist also wrote that in business, intelligence is mostly judgment. I think that supports this reading more than his own. When intelligence gets cheap for everyone, the scarce thing is knowing what to do with it. That is where profit will go, to the few companies that use the saving for something rivals cannot copy quickly, like a promise to answer within an hour that only their own customer history makes possible. We have said that the large gains come when companies redraw the work instead of speeding up old steps. Redrawing the work is also what lets a company keep part of the gain.
Where you will see it first
Expect the effect first where the work is routine writing and paperwork and where customers can compare easily: insurance quotes, bank support chats, airline rebooking, tax preparation. Margins in those businesses may stay the same for years while prices and waiting times keep falling. For an employee, the pressure will mostly arrive as a higher expectation of how fast things get done, not as a share of the saving.
For anyone who runs a team or a business, the useful test is simple. Pick an AI tool you pay for and ask whether customers would notice within a month if you switched it off. If they would, it is paying for itself by keeping you level with rivals, even when your margin does not move. If they would not, it is a cost, however modern it feels.