Picture a planner at a regional parts distributor, fifty-two years old, twenty-one years in the same warehouse system. This is an invented person, a stand-in for the kind of worker the data below describes. In March her team’s purchase-order matching goes to a software agent, and her role is listed as eliminated. She sends forty applications in the first month. The replies ask whether she has experience with the new tooling, which she is, in the narrow sense, the person who has been correcting its output all spring.

Her case is made up, but the question it raises was put to official data last week. Goldman Sachs economist Pierfrancesco Mei looked at what happens to people who lose jobs that AI can do, and Fortune reported the note on 30 October. Workers displaced from AI-exposed jobs since 2025 took about a month longer to find new work than other displaced workers, and about 10 percentage points more of them were still unemployed a year later. The note calls the newest group one that “likely captures the first wave of AI-displaced workers.”

What Goldman measured

The data come from the Current Population Survey’s Displaced Worker Supplement, which the Census Bureau runs every two years. It asks people who lost a job involuntarily in the previous three years what happened next. Goldman analysed more than 15,000 workers who lost jobs between 2016 and 2025, including new data on 2023 to 2025.

No survey asks whether AI caused a layoff, so Goldman built a proxy. Each job loss was scored on the occupation’s exposure to AI, the adoption of AI in the industry and whether the position was eliminated. The top 30 percent were treated as likely AI-displaced. That label is Goldman’s construction, not a worker’s own account, and the result depends on it.

The gap by age is wide. Workers 35 and older took about five weeks longer than other displaced workers to return to full-time work; those under 35 took about three. Goldman found no significant gap between the two groups for job losses before 2023, which is the check that matters: it suggests the score picks up something that changed when AI adoption rose, not a permanent difference between the occupations. The note also says rehired workers, older ones especially, often land in roles with similar or higher AI exposure. Mei says younger workers “may be able to adapt more flexibly to AI disruption,” and Goldman adds that it is “still too early to assess the impact along all of these dimensions.”

An older pattern

The finding repeats one Goldman published in April. Mei and her colleague Jessica Rindels used more than 20,000 people in the National Longitudinal Surveys, sponsored by the Bureau of Labor Statistics, to follow workers whose occupations shrank because of technology in each decade since 1980. Over the ten years after a job loss, their real earnings grew nearly 10 percentage points less than for workers never displaced, and 5 points less than for other displaced workers. They also took about a month longer to find a new job.

The mechanism in that note is occupational downgrading: displaced workers moved into more routine roles that asked for fewer analytical and interpersonal skills. Younger, college-educated, urban workers lost about half as much over the decade, because they changed occupations more readily and moved into more analytical work. Workers who entered vocational or technical programmes within three years of displacement did better, with a probability of returning to unemployment 10 points lower. In that note Goldman also estimated that AI could displace 6 to 7 percent of US workers over the next decade, an estimate and not a count.

The October note therefore says less that AI is new than that its first casualties are being treated like the casualties of earlier technology, and with similar results.

What the official counts show

The Bureau of Labor Statistics published its own Displaced Workers survey on 27 August. It counted 7.4 million workers displaced from 2023 to 2025, of whom 3.3 million had held their jobs for at least three years, 746,000 more than in the previous survey. By January 2026, 66.1 percent of those long-tenured workers were working again, roughly the same as two years earlier. The rate was 72.9 percent for ages 25 to 54, 57.3 percent for ages 55 to 64 and 38.6 percent for those 65 and older.

Pay is where the change shows. Among those who were back in full-time work, about 49 percent earned as much as or more than at the job they lost. Two years earlier the share was about 62 percent. The bureau does not attribute the drop to AI, and Goldman’s score is its own, but the two sources read together describe a market where getting a job back has become easier than getting the same pay back.

Employers are saying more about AI. The outplacement firm Challenger, Gray & Christmas reported that employers cited AI in 120,136 announced job cuts through September, about 21 percent of all cuts and the top stated reason, according to Fortune’s account. An announcement is a company’s explanation, and the cause is not audited.

Where the evidence disagrees

None of this shows that AI is emptying the labour market, and the studies Fortune sets beside the note say so. The Yale Budget Lab, using Current Population Survey data as Fortune reports, found that AI exposure, automation and augmentation measures “show no sign of being related to changes in employment or unemployment.”

The Stanford Digital Economy Lab reaches a narrower conclusion with payroll data from ADP. In a revision published on 12 August, Erik Brynjolfsson, Bharat Chandar and Ruyu Chen report that employment of 22- to 25-year-olds in the most exposed occupations is about 19 percent below where it would be had it tracked less exposed jobs. The gap was 15 percent in the July 2025 data. From November 2022 to June 2026, employment in the two most exposed groups for that age fell about 11 percent, while the three least exposed grew about 10 percent. Yet the authors write: “We do not see widespread, economy-wide job displacement associated with AI,” and “Experienced workers show no comparable gap.” They call their patterns “descriptive patterns, not causal estimates of the effect of AI,” note that the gaps shrink when education is accounted for, and note that some differential trends predate generative AI. The adjustment appears mainly as fewer hires, not more firings.

The sources therefore draw two different groups. Stanford finds the young shut out at the door; Goldman and the bureau find the older worn down after the exit. A reader can hold both: the damage may fall at the ends of a career, while the middle shows little. Goldman’s sample also covers only people who had a job to lose, which leaves out those who never got one.

What it means for the planner

The older worker in Goldman’s data and the young applicant in Stanford’s are not the same person, and the two studies do not describe a single trend. What they share is a cost that arrives late and unevenly: a month more of searching, a year more of unemployment for a tenth of the group, a paycheque that no longer matches the old one. Goldman’s April note found one thing that helped: technical or vocational training within three years of the job loss.

Back in the invented warehouse, the planner is in her sixth month of searching. The tooling that replaced her role still needs someone to check what it produced, and a smaller company has just posted that job at a lower salary. She reads the offer twice, calls the number, and asks what hours the shift covers.

References

  • Brynjolfsson, E., Chandar, B., & Chen, R. (2026). No Widespread Displacement, but the AI Employment Gap for Young Workers Has Widened to 19% (Stanford Digital Economy Lab, summarising “Canaries in the Coal Mine? Six Facts About the Recent Employment Effects of Artificial Intelligence”). digitaleconomy.stanford.edu
  • Fortune (2026). Goldman Sachs on AI layoffs, displaced workers and past tech disruptions (30 October). fortune.com
  • Lichtenberg, N. (2026). Goldman job displacement: 40 years of data on scarring (report on the Mei and Rindels note). Fortune, 6 April. fortune.com
  • U.S. Bureau of Labor Statistics (2026). Worker Displacement: 2023-25. bls.gov