AI Labor Tracker / Foundations

Post-labor economics: who loses the work, and who gets it

Machines are taking on more of the paid thinking, and the useful question is no longer whether work changes but who loses it, who gains it, and how long the gap lasts.

Work is not ending. It is moving, and it is moving faster than the statistics we use to follow it.

That is the whole premise of this publication. “Post-labor” sounds like a world with no jobs.

I don’t think that is where we are headed in the next decade, and the data don’t say so either. What the data do say is that a growing share of the thinking people used to be paid for is now done by software, and that the gains are landing in different places than the losses.

So the question we will keep asking is simple. When a machine takes over a task, who loses the work, who gains it, and how long does the gap last?

Start with the upside

The good news is real, and it deserves to go first.

American workers produced more per hour in the second quarter of 2026 than a year earlier, up 2.2%, according to the Bureau of Labor Statistics. Since late 2019, productivity has grown at 2.1% a year. In the previous business cycle, 2007 to 2019, it managed 1.5%.

Output has grown while hours barely moved. Between the first quarter of 2023 and the second quarter of 2026, the BLS output index for the nonfarm business sector rose from 116.3 to 127.1. The hours index went from 105.4 to 105.9 over the same stretch (BLS).

Meanwhile, people adopted the tools at a speed nobody has seen before. Nearly 40% of Americans aged 18 to 64 had used generative AI by late 2024, about two years after ChatGPT launched, according to a survey by economists Alexander Bick, Adam Blandin and David Deming (NBER). Measured from each technology’s first mass-market product, they found it was being picked up faster than the personal computer or the internet.

That is the upside. More output, cheaper knowledge, and tools almost anyone can use. I’ll take it.

Then the labor question

Now read the same numbers from the other direction.

Real hourly compensation fell 0.1% over the four quarters to mid-2026, even as productivity rose 2.2% (BLS). Workers produced more and, after inflation, were paid a little less for each hour.

And labor’s share of output, the slice of business income that goes to workers as pay and benefits, fell to 52.8% in the second quarter of 2026. That is the lowest level since the series began in 1947, according to the BLS. For most of the post-war period it sat between 60% and 66%.

Be careful with that figure. The labor share started sliding after 2000, long before chatbots, and economists still disagree about why. One record low does not prove a machine did it.

But it tells you which direction the money is flowing. If software is doing more of the paid thinking, the labor share is exactly where you would expect it to show up first. That is why it sits at the top of our dashboard.

Let’s talk about the scary parts

The fear most people carry is that a whole category of work vanishes overnight. The public is clearly worried.

In August 2026, 71% of American adults told Pew Research Center that AI will lead to fewer jobs over the next two decades, up from 64% in 2024. Gallup found 27% of US workers now worry technology could make their own jobs obsolete, a record and roughly double the 13% it recorded in 2017.

So what does the data show so far?

Overall, the job market is soft but not collapsing. Employers added 29,000 jobs in September 2026 and unemployment was 4.2%, inside a 4.1% to 4.3% range it has held since March (BLS).

Indeed’s job postings index stood almost 4% above its February 2020 level in mid-October (Indeed Hiring Lab). That is not the picture of mass replacement.

The damage, where it shows, is concentrated. Using payroll data from ADP through June 2026, Stanford’s Erik Brynjolfsson, Bharat Chandar and Ruyu Chen found that workers aged 22 to 25 in the jobs most exposed to AI had employment 19% below where the trend would have put them (Stanford Digital Economy Lab).

Experienced workers in the same jobs showed no gap. The authors call their findings descriptive, not proof of cause.

That is the pattern I worry about most. The machine doesn’t fire the senior person. It quietly removes the first rung of the ladder that the next senior person would have climbed.

The other honest answer to the scary part is history. A team led by MIT’s David Autor found that “the majority of current employment is in new job specialties introduced after 1940” (NBER). Most of the jobs people do today did not have a name 80 years ago.

New technology has always destroyed some work and created more. The open question is whether that holds when the technology does thinking rather than lifting.

What “post-labor” means here

The phrase “post-labor economics” has been circulating online for a few years, and much of the current conversation runs through the writer David Shapiro, who argues that ownership needs redesigning as labor’s value falls. We borrow the phrase, not a conclusion.

For this publication, post-labor economics means the study of an economy in which a rising share of output comes from machine intelligence rather than human hours. It covers what happens to pay, to jobs, to communities and to the dignity people get from supporting themselves. It is a description of a direction, not a prediction that work disappears.

In my opinion, the word that matters most in that definition is “share.” Nobody needs every job to vanish for this to be the economic story of our lives. It is enough for the share of income going to work to keep falling while the share going to machines and their owners keeps rising.

Economists disagree sharply on how far this goes. MIT’s Daron Acemoglu estimates AI will add at most about 0.66% to total factor productivity over ten years (NBER). Goldman Sachs expects about 9% of US workers, roughly 15 million people, to be displaced or reallocated over a decade (Fortune).

The IMF says 40% of jobs worldwide are exposed to AI, and 60% in advanced economies (CNBC). At the far end, papers by Anton Korinek (NBER) and Pascual Restrepo (NBER) model what happens to wages if machines can eventually do nearly every task.

Those are not small differences. That range is the reason to measure instead of guess.

What we will track

This publication will follow seven things, every month, with the source linked every time.

Technology leaps of the past. Fire, steam, electricity, the car, flight and the internet each wiped out some work and created more, on different timelines. Our next piece tests whether AI is bigger than the Industrial Revolution against that record, including how long each transition hurt (Bigger than the Industrial Revolution?).

Jobs cooling and heating. Which occupations are losing postings, employment and pay, and which are gaining. Our job thermometer uses BLS employment counts, BLS projections and Indeed postings. A lot of the heat right now is in trades that don’t need a four-year degree, which I find encouraging.

Macro impacts. Productivity, labor share, hours, real pay, unemployment and AI investment. These are the numbers in Figure 1, and they get updated as BLS publishes. The hardest macro question, who pays for people if wages shrink, gets its own piece.

The online conversation. What people say about AI and jobs, and whether the claims going viral are real, current or recycled. A lot of what spreads as “breaking” turns out to be a quote from 2025. We will say so.

Science fiction’s predictions. Writers have imagined this for a century. In 1930 John Maynard Keynes warned of “technological unemployment,” which he defined as “unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can find new uses for labour” (Keynes, 1930). He expected that “three-hour shifts or a fifteen-hour week” might be how we would share the remaining work. Americans on private payrolls averaged 34.4 hours a week in September 2026 (BLS). In 1964 Isaac Asimov wrote that “mankind will therefore have become largely a race of machine tenders” (New York Times). Supervising AI agents is starting to look a lot like machine tending. We score these predictions against what actually happened in the sci-fi scorecard.

Real economists. Who argues what, with their actual numbers and links, from skeptics like Acemoglu to the scenario builders. When we quote an economist, the quote will be sourced.

The Gross Intelligence Product. Our own index of how much useful machine thinking the world produces, built entirely from other people’s published data: electricity used by AI data centres, chip efficiency from Epoch AI, and model quality from the Artificial Analysis Intelligence Index. The name was first proposed by Lee Seung-hyun of the Korean AI company 42Maru in his 2026 book AI Wealth of Nations (Seoul Economic Daily). Our method is introduced in Introducing the Gross Intelligence Product. Set against hours worked, it gives us one way to see whether machine output is replacing paid work or adding to it.

How we will work

Every factual claim on this site links to its primary source. When a number can’t be verified, we cut it or say so.

When the data are an estimate, we label them. When we are wrong, the correction goes at the top.

We are not partisan. The labor question doesn’t belong to one party, and both have reasons to want the numbers to say something they don’t.

And we will keep this relevant to the person reading it: the task that moved to software, the role that changed around it, and the paycheck that did or didn’t follow. That’s the level where this story actually lives.

Fair warning: this publication will contain more charts than jokes.

Where this leaves you

Knowledge is free. Creativity isn’t. That line gets a full piece in this series.

The more machines handle routine thinking, the more the market pays for judgment, taste, trust and hands-on skill. Some of the best-paid growth in the country right now is for electricians, line workers and HVAC technicians, jobs you can enter through an apprenticeship rather than a degree. We cover that in detail in the job thermometer and in our profile of the data center electrician.

If you are early in your career, the Stanford data say you are the one most exposed, so act like it. Use the tools every day until you are the person on your team who knows them best.

Pick work where the human part is judgment or hands, not routine. And watch the numbers we publish here, because the shift will show up in the data before it shows up in your inbox.

Lean in, keep learning, and measure what is actually happening rather than what is being shouted about. That’s my recommendation.

Sources

1. U.S. Bureau of Labor Statistics, “Labor share at its lowest level, 52.8 percent, in second quarter 2026,” The Economics Daily, Sep 15, 2026. https://www.bls.gov/opub/ted/2026/labor-share-at-its-lowest-level-52-8-percent-in-second-quarter-2026.htm
2. U.S. Bureau of Labor Statistics, “Productivity and Costs, Second Quarter 2026, Revised,” Sep 3, 2026. https://www.bls.gov/news.release/prod2.htm
3. U.S. Bureau of Labor Statistics, “The Employment Situation, September 2026,” Oct 2, 2026. https://www.bls.gov/news.release/empsit.htm
4. Bick, Blandin and Deming, “The Rapid Adoption of Generative AI,” NBER Working Paper 32966. https://www.nber.org/papers/w32966
5. Brynjolfsson, Chandar and Chen, “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,” Stanford Digital Economy Lab, revised Aug 12, 2026. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/
6. Autor, Chin, Salomons and Seegmiller, “New Frontiers: The Origins and Content of New Work, 1940-2018,” NBER Working Paper 30389. https://www.nber.org/papers/w30389
7. J. M. Keynes, “Economic Possibilities for our Grandchildren,” 1930, Marxists Internet Archive transcription. https://www.marxists.org/reference/subject/economics/keynes/1930/our-grandchildren.htm
8. Isaac Asimov, “Visit to the World’s Fair of 2014,” New York Times, Aug 16, 1964. https://archive.nytimes.com/www.nytimes.com/books/97/03/23/lifetimes/asi-v-fair.html
9. Daron Acemoglu, “The Simple Macroeconomics of AI,” NBER Working Paper 32487. https://www.nber.org/papers/w32487
10. Anton Korinek and Donghyun Suh, “Scenarios for the Transition to AGI,” NBER Working Paper 32255. https://www.nber.org/papers/w32255
11. Pascual Restrepo, “We Won’t be Missed: Work and Growth in the AGI World,” NBER Working Paper 34423. https://www.nber.org/papers/w34423
12. Fortune, “AI could displace one in 11 workers, Goldman says, but the layoffs are waiting for a recession,” Oct 22, 2026. https://fortune.com/2026/10/22/goldman-sachs-forecast-how-many-jobs-lost-to-ai-9-percent/
13. CNBC, “IMF warns of AI ‘tsunami’ for job seekers,” Oct 12, 2026. https://www.cnbc.com/2026/10/12/imf-world-bank-ai-jobs-inflation-.html
14. Pew Research Center, “Young adults in the US are increasingly wary of AI, concerned it will take jobs,” Aug 18, 2026. https://www.pewresearch.org/short-reads/2026/08/18/young-adults-in-the-us-are-increasingly-wary-of-ai-concerned-it-will-take-jobs/
15. Lydia Saad, “More U.S. Workers Fear Losing Their Jobs to Technology,” Gallup, Sep 15, 2026. https://news.gallup.com/poll/714368/workers-fear-losing-jobs-technology.aspx
16. David Shapiro, “Post-Labor Economics” (Substack section), 2026. https://daveshap.substack.com/t/post-labor-economics
17. Seoul Economic Daily, “AI emerges as national strategic asset, new book argues,” Mar 22, 2026. https://en.sedaily.com/technology/2026/03/22/ai-emerges-as-national-strategic-asset-new-book-argues
18. Indeed Hiring Lab, “US Labor Market Snapshot: October 2026,” Oct 22, 2026. https://hiringlab.indeed.com/2026/10/22/us-labor-market-snapshot-october-2026/