AI Will Take Its Jobs in the Next Recession
B · Useful, limited Labour economist Gad Levanon argues from BLS history and published research that AI job losses will arrive in the next recession; a sourced and testable argument, but a forecast in a newsletter, not new analysis.
Summary based on the full text·Original: about 7 minutes to read
In short
In his Labor Matters newsletter (16 October 2026), labour economist Gad Levanon argues that the current calm about AI and jobs repeats the consensus on manufacturing in 1999-2000. Then, firms held a stock of labour-saving options they did not use while profits were good, and cut during the 2001 and 2008 recessions. He predicts AI-exposed office jobs will go the same way in the next downturn, and many will not return.
Key numbers
- 17.3 million to 11.5 millionUS manufacturing employment, 2000 to 2010
- 88 percentshare of the manufacturing job loss that fell in two recession windows
- about 1.6 million jobs, or 4 percentFIIPB white-collar sectors below their 2005-2019 trend by September 2026
- 1.5 to 2 percentLevanon's estimate of knowledge workers in jobs that would not exist without AI
As reported by the source.
English·Web page·opens gadlevanon.substack.com
What it reports
- Manufacturing peaked at 17.6 million in March 1998 and had slipped only about 300,000 by mid-2000. BLS projected 3% growth for 2000-2010; employment fell from 17.3 million to 11.5 million, a loss of 5.75 million.
- Losses ran at 76,000 a month in 2001-04 and 86,000 a month in 2007-10, against about 12,000 a month in the 2004-07 expansion: 88% of the loss sat in the two recession windows.
- Jaimovich and Siu: 88% of routine-job decline since 1990 occurred around recessions; manufacturing was only 37% of it.
- Autor, Dorn and Hanson: the most China-exposed areas had not recovered employment rates even by 2019. Pierce and Schott: lasting rises in deaths from suicide, drugs and alcohol in PNTR-exposed counties.
- Finance, insurance, information and professional and business services (FIIPB) peaked in April 2023 and by September 2026 were about 1.6 million jobs, or 4%, below trend: the first non-recession decline in decades.
- AI also creates jobs: he estimates 1.5-2% of knowledge workers hold AI-dependent roles, plus data centre and power buildout.
How they got it
Historical BLS series, published academic studies and his own sector tracking. The explanation for recession timing is his reading, as he states.
Limits
- A forecast by analogy; AI adoption may follow a different path than offshoring and factory automation.
- The FIIPB trend gap depends on the chosen 2005-2019 baseline and does not isolate AI as the cause.
- He discounts studies finding small AI effects as flawed but does not show why here.
Why it matters for companies
If the stock of AI-enabled cuts builds quietly, workforce plans that assume gradual change will be wrong-footed in a downturn, when cuts happen fast and under cover of cost pressure. Firms can use the current period to redesign roles, retrain and build internal mobility, rather than making all the decisions at once in a recession. The weak position of young graduates is already a warning sign for entry-level pipelines.
Summary based on the full text. Read the original for the full argument and data.Read the original on gadlevanon.substack.com
- Tags
- Workforce, Displacement