I have covered enough AI-and-jobs stories to expect one version of events: the machine gets better, the headcount gets smaller, the survivors get worked harder for the same money. So it is worth sitting with a finding that cuts against that story. PwC's 2026 Global AI Jobs Barometer, built on more than a billion job ads across 27 countries and territories, sorts companies into AI exposure quartiles using sector tagging and the occupational tasks in their job postings. By that measure, the companies most exposed to AI are not shedding people. They are hiring more of them, and paying more too.
Since 2018, headcount at the most AI-exposed companies has grown 52 percent, against 36 percent at the least exposed. Productivity, measured as revenue per employee, which PwC calls turnover, grew about 40 percent faster in the same group. Wages rose 24 percent versus 17 percent. Joe Atkinson, PwC's global chief AI officer, frames it as a fork in the road: "The companies seeing the greatest returns on AI are using it to amplify human expertise, accelerate innovation and create entirely new sources of value. As a result, they are pulling further ahead on productivity and growth than companies that focus primarily on automation."
That is the headline PwC wants you to carry away, and on the numbers above it holds. But stay with the report past the headline and it tells a second story, one PwC put in its own findings deck without quite saying it out loud.
Inside the most AI-exposed quartile, PwC isolated the top 20 percent of companies by productivity growth and called them the superstars. Their productivity grew 163 percent since 2018, nearly five times the average for AI-exposed firms as a whole. Their wage growth ran to 68 percent, against 24 percent for the exposed quartile overall. Their headcount growth, though, was 44 percent, a figure PwC states outright in a callout on the headcount chart of its findings report, page 13, only slightly below the 52 percent recorded for the whole exposed group. Read that again: the companies capturing almost all the productivity gain and almost all the wage gain are not the ones doing a disproportionate share of the hiring. The hiring is spread fairly evenly across every AI-exposed firm that made the cut. The money is not.
That is the actual shape of the "superstar effect," and it matters because it answers the question the lead figures dodge. Do AI's gains concentrate or spread? Both, but not in the same place. Headcount spreads. Pay and productivity pool at the top fifth. If you work at an AI-exposed company, the odds that firm is adding jobs rather than cutting them have risen across the board, not just at the very top. Whether you are one of the people whose wage grew 68 percent or one of the people whose employer merely says the average is up 24 percent is a separate question this data cannot answer, because it is reported at the company level, not the worker level.
PwC's own findings report admits a further wrinkle: the analysis only includes firms with financial data in both 2018 and 2024 or 2025, which means every company that closed, shrank below the threshold, or got bought out along the way is invisible. "The key interpretation is therefore the relative difference in headcount growth between more and less AI-exposed companies," the report says, "rather than the absolute growth rate compared with the broader economy." That is a careful way of saying: this measures who survived and won, not what AI does to a labor market as a whole.
None of that makes the top-line finding false. It makes it partial, the way most good news about work usually is. The companies exposed to AI are hiring, and they are paying more on average. The honest addition is the question I keep asking of every number like this: who benefits, and who is just carrying the average on their back. That average is doing a lot of the same work averages always do: hiding who actually cashed in.
