September added 29,000 jobs, and AI shows up in hiring plans, not layoffs

The US added 29,000 jobs in September, about a third of the 90,000 economists expected, and the unemployment rate rose to 4.2%. Hours later Claudia Sahm told Bloomberg that an economy growing this fast with this few new workers is a puzzle, and that AI is not yet big enough to solve it. The executives who actually run AI inside their companies told a different story the same week. Nobody is firing people because of it. They are just not hiring the next one.

Sahm's puzzle is growth without workers

The numbers in the October 2 report were weak in a quiet way. July and August lost 60,000 jobs to revisions. Average hourly earnings rose 0.1%. Healthcare added 17,000 roles and construction 11,000, and finance shed 7,000. Layoffs stayed low. Sahm called it the low hire, low fire labor market and said it has now run for at least three years. What bothers her is that output keeps rising anyway. Tom Keene put nominal growth north of 6% when he asked her how that adds up.

So right now if labor force is not growing and the economy, the activity is growing, it's got to come from productivity. But like, you know, AI adoption is not widespread enough to explain all of this. So, it's curious. It really is.

Claudia Sahm, October 2, 2026, source

In her instant reaction after the print she went further. "Things are out of sync in a way that just doesn't feel like it can keep going, and yet it has," she said, before adding the line that should worry anyone who is graduating this year. "This is not normal."

Most of the hiring drought is demographic

Before blaming software, take the boring explanation seriously. Sahm puts most of the slowdown on the supply side. Immigration fell sharply, some temporary visas were revoked, and baby boomers keep retiring. A labor force that is barely growing does not need 150,000 new jobs a month to hold the unemployment rate steady. She also flagged a distortion that runs the other way.

I think you can go to a few immigrant-intensive industries, like maybe in construction, and you see some wage growth pick up, which could be labor shortages. But, oh yeah, we're also building a bunch of data centers, so that's pushing up construction.

Claudia Sahm, October 2, 2026, source

So AI shows up in this report, but as a source of demand for electricians and concrete crews, not as a job killer. That fits what one big employer said in Indiana.

Lilly mapped its jobs and found parts, not wholes

Dave Ricks took an audience question at the Economic Club of Indiana about what AI will do to Lilly's staff. He spoke on September 22 and the video went up October 2. His answer is the most useful corporate data point I found this week, because Lilly says it actually did the work.

We've done a lot of mapping of this, and I think interestingly there aren't that many like whole jobs that have been replaced. There's a couple examples, but not too many. Mostly it's like parts of everyone's job.

David Ricks, September 22, 2026, source

His real constraint is the opposite of an AI surplus. Lilly's site in Lebanon, Indiana has "about 8,500 building trades people there today working," he said, and "there's only 17,000 certified in the state." Workers drive in from Iowa and Missouri and live in trailers for six months. Ricks said those jobs pay $250,000 a year. That is Sahm's construction point with a name and a number attached.

Where AI does bite is the job that never gets posted

The companies furthest into AI describe something the payroll survey can't see. Nobody gets fired. The requisition just never opens. Guillermo Rauch, who runs Vercel, explained it with a sponsorship deal on a Slovenian podcast recorded around September 27 and published October 1.

So if I sign a partnership with Mercedes, okay, I have to hire another seven designers because we used to live in this linear world. We're going into a nonlinear world.

Guillermo Rauch, September 27, 2026, source

His design team now builds the tools that make the posters rather than the posters themselves. Seven designers who would once have been hired never get an offer, and no unemployment office ever hears about it. Mike Novogratz told Anthony Scaramucci on October 1 that he sees the same thing, and he was careful to say it is still narrow. "We've seen them in some areas only, right? The young coders are not getting hired," he said, and described a founder who has gone "from 55 to 13 engineers and his company's bigger."

Novogratz also hedged. "We haven't seen job losses due to AI yet and maybe we never will," he said in the same answer. One founder's anecdote is not a macro trend. A software company is the easiest place there is to swap headcount for tokens, and Vercel sells the tools that do it.

Cathie Wood reads the same tape as a boom

Cathie Wood spent her October 2 jobs-day webcast arguing that rising real yields mean the economy is working, and that AI will push productivity growth to 5% or 6%. She did not pretend the youth numbers look good.

But do look at the purple line that's the 16 to 24 year old age cohort. That is back at 9.9%.

Cathie Wood, October 2, 2026, source

Her advice to those graduates was to start a company with AI and no employees, which is a strange thing to offer as comfort. She also guessed that the soft 0.1% wage gain came from "a lot of technology jobs tailing off," since those are the high earners. I can't check that from the headline release, and she offered it as a maybe. But it points the same way as Rauch and Novogratz. The first group to feel AI is young, white collar and technical, and it feels it as a closed door, not a pink slip.

The 11 million estimate and a Fed that is looking elsewhere

McKinsey put a size on the long run in a report released September 29. About 11 million US workers, roughly 7% of the workforce, may need to change occupations by 2035, with a range of 6 million to 16 million. That implies about 645,000 career switches a year against a historical 215,000. The groups most exposed are office support, retail and sales, and transportation, which is a different list from the coders and designers in the interviews above. Novogratz thinks the political fight starts with next month's midterms.

When you've got rising unemployment, we're going to be in a recession cutting rates, not hiking rates.

Mike Novogratz, October 1, 2026, source

That is a forecast about some future year. Today the Fed is doing the opposite. It raised rates a quarter point in September, to 3.75% to 4%, its first hike since 2023, and Kevin Warsh said inflation "is too high, and has been for too long." Sahm said the September report gives the Fed little new information before the October 27 to 28 meeting. A labor market that is merely stuck does not move a central bank chasing energy inflation. Reid Hoffman, on the Newcomer podcast on September 30, gave the honest version of everyone's uncertainty.

And by the way, no one actually in fact really knows. Neither the people who go, oh my god, it's going to be huge in a year or two, or the people who are like, oh no problem, nothing this year.

Reid Hoffman, September 30, 2026, source

So is AI the reason hiring stopped?

Mostly no, and partly yes. The bulk of the slowdown is fewer workers arriving, through immigration and retirement, and Sahm is right that AI is too thin to explain an economy growing this fast. But the AI effect is real at the margin, and it is aimed at exactly the people the survey can't see, the graduate who never gets an interview. Ricks says it takes parts of jobs. Rauch says it takes the next hire. Neither shows up as a layoff. If you want an early signal, skip the headline payroll number and watch the 16 to 24 unemployment rate and the finance line, which has lost 129,000 jobs since May 2025. I would also stop expecting the Fed to react. It won't until unemployment moves, and this kind of displacement is built not to move it.