Goldman Sachs says AI layoffs may leave lasting scars on workers

Goldman Sachs published research on 30 October 2026 arguing that job losses linked to artificial intelligence are beginning to leave a lasting mark on workers. According to Fortune's report on the note, the research compares the early AI wave with four decades of technology-driven layoffs and finds a familiar pattern. The author, Goldman economist Pierfrancesco Mei, describes the effects from earlier technology waves as long-lasting and says the first AI-displaced group shows comparable effects.

What the research found

The analysis covers more than 15,000 displaced workers who lost jobs between 2016 and 2025, based on a US Labor Department survey. Goldman had no survey question that asks whether AI caused a job loss. So it built a score using how exposed an occupation is to AI, how far AI has spread in the worker's industry, and whether the position was eliminated. It treated the top 30% of that score as likely AI-displaced. Goldman says the 2023 to 2025 data probably captures the first wave of AI displacement.

Workers displaced from AI-exposed jobs since 2025 took about one month longer to find new work than other displaced workers. They were also roughly 10 percentage points more likely to still be unemployed a year after the layoff. Among workers aged 35 and over, the gap to return to full-time work was about five weeks. Younger AI-displaced workers returned to full-time work faster, cycled through fewer temporary jobs, and were less likely to hold a second job alongside a new full-time role.

The note also points to the age gap in the government figures for all laid-off workers, not only those in AI-exposed roles. That suggests age is a factor in recovery regardless of the cause of the layoff.

How the AI figure should be read

The employer figure is the one most readers will repeat, so it is worth being exact. According to Challenger, Gray & Christmas, employers announced 573,195 US job cuts through September 2026, and 120,136 of them cited AI, which is about 21% of the total. Challenger's count records the reason an employer gives in its announcement. It is not a measure of how many people lost jobs because of AI, and announced cuts are not the same as completed layoffs.

Goldman's own forecast is larger and further out. In an October 2026 update reported by Fortune, Goldman estimated that up to 9% of US workers could be displaced during the AI transition, which it puts at roughly one in 11 workers, with slower hiring adding a further drag of about 3%. Goldman also suggests the largest layoffs may come during a recession rather than now. Stanford's Digital Economy Lab has taken a different view. Its researchers say they do not see widespread, economy-wide displacement from AI, and they link the employment gap for workers aged 22 to 25 in the most exposed jobs mainly to reduced hiring.

Why this matters for a business that automates work

If you automate data entry, reconciliations, billing follow-ups or reporting, the people in those tasks are the ones Goldman is describing. The research suggests that a role removed by automation can cost a worker several weeks of income and, for some, a lasting drop in earnings. Earlier Goldman work, using data back to 1979, found that displaced workers' real earnings grew more slowly for years after the job loss. That is a reason to plan changes rather than make them abruptly.

The findings also say something about hiring. Goldman's data suggests reduced hiring is part of the picture, not only layoffs. A business that stops filling a role because a workflow now runs by software should count that as a decision with the same human weight as a dismissal.

What changes in practice, at any size

A five-person firm and a group with thousands of staff face the same basic questions. The first is which tasks are routine and repeatable, and which need judgement. The second is who does the work today and what they could move into. The third is what a person still checks before anything leaves the building.

Where a role is reduced, the rules on notice, consultation and severance depend on the employer's location and the contract. In the United States those rules vary by state and by whether a federal law such as the WARN Act applies. This should be agreed with the employee or checked with an employment lawyer before any change is announced.

The limits of automation

Automation is not a fit for every task. Work with unclear rules, frequent exceptions or heavy customer judgement usually needs people. Software that runs on bad or incomplete data will make the same errors faster. An automated workflow also needs an owner who watches it, updates it when rules change, and can explain its output to an auditor or a customer. Results depend on your own volumes and error rates, which is what a free audit measures before anything is built.

What to watch next

Three things will show whether this research changes the picture. The first is Challenger's next monthly report, which will show whether AI-cited announcements return after a quiet September. The second is the Labor Department's displaced worker data for later years, which Goldman says will test whether the first AI wave is widening. The third is whether a recession arrives, since Goldman's argument is that AI-related layoffs could grow sharply in a downturn. For a business, the practical step is to map which tasks are automated, who does them, and what happens to those people before the next change.