Picture a payroll manager at a regional insurer, a woman who does not exist, on a Friday afternoon at the end of October. She has a spreadsheet open with three columns: headcount, claims handled, pay budget. Headcount has not moved in a year. Claims handled are up, partly because the company’s new software sorts the simple ones without a person. The pay budget for January allows three per cent, which is what the budget allowed last year and the year before. She has no way of knowing that, across the country, the figure she is protecting has been losing ground to the figure above it in the company’s income statement.

The country-sized version of her spreadsheet came out at the end of the month. According to the Bureau of Labor Statistics’ labor productivity and costs report, as Fortune read it on 31 October, labour’s share of US output fell to 52.8 per cent in the second quarter of 2026. That is the lowest since the series began in 1947. The previous quarter was 53.6 per cent, and the average for this decade has been about 55.6 per cent. The series can be checked directly. The BLS website refuses automated requests for the release, but its time-series pages are open, and they carry the same series as an index with 2017 set to 100.

What the series shows

The index sat close to 100 through the six years from 2016 to 2021, apart from a pandemic spike to 104.2 in the second quarter of 2020. It was 99.1 in the first quarter of 2022. Then it fell, to 97.5 in the second quarter and 96.9 by the end of that year, and it stayed around 96.2 to 96.9 through 2023 and 2024. A second slide followed: 95.7 at the end of 2025, 94.8 in the first quarter of this year and 93.4 in the second, a figure BLS marks as revised. From the start of 2022 to now the index has dropped 5.7 per cent. About two fifths of that happened in 2022 alone.

The same tables show what moved underneath. Real hourly compensation, the pay a worker gets per hour after prices, fell 1.4 per cent in the first quarter of 2026 and 3.3 per cent in the second, while output per hour rose 0.8 and 1.4 per cent. Hours worked rose by 0.7 and 0.3 per cent. The first slide, in 2022, looked different: real hourly compensation fell 6.7 and 8.6 per cent in the first two quarters, once prices are counted, while output per hour was falling as well. In neither episode do the tables show hours collapsing. They show pay per hour giving ground.

Who points at automation

Fortune quotes Raymond Robertson, a labour economist at Texas A&M’s Bush School of Government. A fall in labour’s share, he says, “has got to be either falling earnings or falling numbers of people”, and he ties the shift toward capital to automation, which he says is replacing workers while output per worker keeps rising. He expects unemployment, which ticked up to 4.2 per cent in September, to rise in the coming months and to keep wages down. The jobs figures give his reading some support. The economy added 29,000 jobs in September, roughly a third of what economists projected, and 584,000 in 2025 against 2 million in 2024.

Fortune is careful about what the figures do not say. It notes that job displacement from AI “has not shown up meaningfully in any economic data”, and it cites competing explanations. Morgan Stanley economists have called the recent productivity rise an “open question” that may be cyclical rather than a product of AI. A research brief from Oxford, cited by Fortune, suggested that some companies dress up layoffs after over-hiring as AI-driven, while automation-related cuts have not yet happened at scale. Mark Regets of the National Foundation for American Policy argues that the immigration crackdown has shrunk the workforce, with a 1.1 million to 1.4 million drop in the foreign-born labour force since January 2025 by his organisation’s count.

The long decline, and what AI studies find

Falling labour share is not new. Loukas Karabarbounis and Brent Neiman documented in 2013 that the global labour share had fallen since the early 1980s, in most countries and industries, and attributed roughly half of it to the falling relative price of investment goods, which pushed firms to substitute capital for labour. David Autor and co-authors argued in 2017 that the cause lay in the rise of “superstar firms”, whose high profits and low labour shares come to dominate their industries. Neither paper is about generative AI. Both are reminders that the series moves for reasons that have little to do with any one technology.

The studies that look for AI directly find something narrower. Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, using ADP payroll data on millions of workers through June 2026, find no evidence of widespread, economy-wide displacement. They do find that employment of 22 to 25-year-olds in AI-exposed occupations is 19 per cent below where it would be had it tracked their less-exposed peers, working mainly through reduced hiring, and that adjustment came through employment rather than base pay. They call these early descriptive indicators, not causal estimates. The Federal Reserve’s October Beige Book reports that five districts had firms reducing staff through attrition, delaying hiring or expecting less job creation after raising productivity with AI.

On the productivity side, the evidence is thinner than the hope. A report from the University of Manchester’s Productivity Institute, commissioned by the IT services firm Bytes Technology Group and summarised by The Register on 30 October, cites a study of AI development tools in which coding activity rose about 240 per cent while software releases rose about 30 per cent, because people still check and test the work. Its lead author, Bart van Ark, says: “The ambition gap lies between those individual efficiencies and sustained business results.” Torsten Slok of Apollo wrote on 24 October that revenue per worker and profit margins are flat for the S&P 493, the 493 companies outside the seven largest technology firms, and that there are “still no signs of AI boosting productivity growth”.

Two readings of one number

The pieces do not fit neatly. Robertson’s account has automation replacing people while output per worker rises, yet output per hour grew 1.4 per cent last quarter, and the Manchester report and Slok both say AI’s gains are not reaching firms’ results. If AI is doing little, something else drove real pay per hour down for two quarters while output rose, and the rival explanations above each have a claim. Robertson himself concedes that “Data right now is very mixed”. The record is also one quarter, and BLS has already revised it once.

What the number does establish is the direction of the money. Output per hour has grown; the pay attached to each hour has lagged it. Whether AI deserves a share of the blame will take years of data and better measures than a single national ratio.

The invented payroll manager will write her three per cent into the January plan, and prices will decide what it buys. Her claims software will keep sorting the simple files. Nothing in the national series says whether that software is why her colleagues’ pay buys less, or only sits in the building while the cause lies elsewhere.

References

  • Autor, D., Dorn, D., Katz, L. F., Patterson, C., & Van Reenen, J. (2017). The Fall of the Labor Share and the Rise of Superstar Firms. NBER Working Paper 23396. nber.org
  • Brynjolfsson, E., Chandar, B., & Chen, R. (2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab. digitaleconomy.stanford.edu
  • Bureau of Labor Statistics (2026). Major Sector Productivity and Costs: Nonfarm Business, Labor share, index 2017=100 (PRS85006173). data.bls.gov
  • Bureau of Labor Statistics (2026). Major Sector Productivity and Costs: Nonfarm Business, Labor productivity (output per hour), percent change from previous quarter (PRS85006092). data.bls.gov
  • Bureau of Labor Statistics (2026). Major Sector Productivity and Costs: Nonfarm Business, Real hourly compensation, percent change from previous quarter (PRS85006152). data.bls.gov
  • Bureau of Labor Statistics (2026). Major Sector Productivity and Costs: Nonfarm Business, Hours worked, percent change from previous quarter (PRS85006032). data.bls.gov
  • Federal Reserve System (2026). Beige Book, October 2026: National Summary. federalreserve.gov
  • Karabarbounis, L., & Neiman, B. (2013). The Global Decline of the Labor Share. NBER Working Paper 19136. nber.org
  • Robinson, D. (2026). AI projects not paying off? Keep at it, says Uni Productivity Institute. The Register. theregister.com
  • Rogelberg, S. (2026). U.S. workers just took home their smallest share of capital since at least 1947, and the lowest ever recorded. Fortune. fortune.com
  • Slok, T. (2026). No Signs of AI Payoff Outside the Tech Sector. Apollo, The Daily Spark. apollo.com