AI Could Displace Up to 9% of Workers, and Goldman Sachs Says a Recession Would Bring It Forward

If AI eventually lifts productivity by 15%, as Goldman forecasts, it could displace up to 9% of workers during the AI transition, with slower hiring weighing on employment by another 3%, in a Goldman Sachs Research note by economist Joseph Briggs released on October 22. In June, Briggs put that share at about 9%, or 15 million workers over a decade. The sharper point is timing, since automation-related disruption is limited when the overall economy is healthy but accelerates during cyclical slowdowns. For Metaintro readers, this guide explains where the numbers come from, which jobs history says are most at risk, why a recession matters, and what help exists if your role is cut.

What Exactly Did Goldman Sachs Say About AI and Jobs on October 22?

The 9% estimate in the new note, which Goldman released on October 22, rests on one forecast. Goldman forecasts that AI eventually lifts productivity by 15% in the U.S. and other developed economies, and on that basis the displacement share works out to roughly one in 11 workers. Briggs went back through decades of data on how technology has replaced workers to study the timing.

Briggs co-leads the global economics team in Goldman Sachs Research, and the note says worry has eased. "Concerns of imminent AI-related labor displacement have cooled in recent months after hitting a fever pitch this spring," Briggs wrote.

"The timing and extent of automation and labor disruption will likely track the overall economic cycle," Briggs wrote.

Where Does the 15 Million Figure Come From?

That headcount figure is older than the October 22 note. Nine percent of workers being displaced by AI would correspond to 15 million workers leaving or being displaced from their positions today and having to find new jobs, Briggs said on a June 2026 Goldman Sachs Exchanges podcast. Goldman's June 25, 2026 Top of Mind research expects that the AI transition could lead to the displacement of over 9% of the labor force, and adds that its prior unemployment-rate estimates likely understate the total amount of displacement if some workers find new positions quickly.

The estimate has climbed. In August 2025, Briggs and economist Sarah Dong put AI displacement at 6% to 7% of the U.S. workforce if widely adopted, within a range of 3% to 14%, using the same productivity assumption. Goldman's new estimates suggest that each 1% increase in productivity due to technological change cumulatively raises the job destruction rate by around 0.5-0.6pp in the two years following the shock, which it calls "roughly 50% higher than our prior displacement estimate" in the June report.

Why Would a Recession Speed Up AI Job Losses?

History is the core of the argument. Historically, Briggs writes, automation-related labor disruption is quite cyclical, with companies tending to automate work and shed labor costs when the economy starts to slow. Companies are often more willing to replace labour with technology when economic conditions deteriorate and pressure to cut costs increases, the Times of India wrote of the note.

The chart behind the claim tracks the share of Americans in routine jobs, such as clerks, secretaries, factory machine operators and mechanics. That share has remained stable during expansions but declined sharply during downturns, and the line drops in steps for the recessions of the early 1980s, 2001, 2008, and 2020. After each of those recessions the line doesn't climb back to where it was.

The damage is also bigger in weak markets. Across 25 occupations disrupted by past technology, employment fell an extra 4% over five years for every point the national unemployment rate climbed. The hit to unemployment from a technology-driven productivity gain was 0.1 to 0.2 percentage points bigger when the overall jobless rate was rising.

Goldman has made a similar point before. In October 2025, Goldman's US economists David Mericle and Pierfrancesco Mei warned that "the full consequences of AI for the labor market might not become apparent until a recession hits." The decline in the routine occupation share has historically mostly occurred in economic recessions, when the economy is less able to create new positions, Goldman wrote in June.

A more concerning takeaway from our analysis is that a cyclical slowdown would likely accelerate AI-related automation and job losses, thereby amplifying the broader business cycle, Goldman economist Joseph Briggs wrote, according to the Fortune report on the note.

Which Jobs Have Been Hit Hardest by Past Waves of Automation?

Briggs looked at 25 jobs hit hard by past technology, including photo processors, movie projectionists, proofreaders, file clerks, and bank tellers. On average, employment in those jobs fell 25% in the five years after it started to decline. Briggs treats those figures as likely upper bounds for AI, because the sample was picked from cases where technology hit hardest.

How the jobs disappeared matters for anyone in an exposed role. Most of the shrinkage came from people losing or leaving jobs, more than from employers hiring fewer people. Separation rates rose in 19 of the 25 occupations, by 3.8 points on average. Hiring rates fell in 13, by 0.7 points on average.

That is why Briggs expects AI-related layoffs to provide a good real-time guide to how much disruption is under way.

Which Workers Look Most Exposed to AI So Far?

The early damage is narrow and concentrated. Employment in call centers, software publishing, management consulting, and advertising services has fallen sharply below trend across the developed world, Goldman Sachs Research found. In that analysis, call center employment stood 39% below trend in the US, 33% below trend in Canada, and 27% below trend in Germany.

Newer workers may carry more of the risk. Recent anecdotes and US labor market data suggest that AI displacement could be stronger among entry-level workers. The wage and unemployment rate gaps between entry-level and experienced workers widened by 1.3% and 0.6 percentage points, respectively, from prepandemic averages.

Entry-level workers in their 20s and 30s, coming into the knowledge and content creation sectors are likely to be most affected by new deployments of AI, Goldman Sachs Research said in March. That is the same group tracked in why some early career workers thrive with AI while others fall behind. Briggs added a caveat in that article, saying "But this is not a foregone conclusion."

The overall effect is still small. A 10% occupational exposure to AI is only associated with a 0.1 percentage point drag to annual headcount growth in France, Canada, and the US, and impacts are limited to a narrow set of industries and workers.

Why Have AI Layoffs Been Limited So Far?

Briggs gives two answers for the gap so far. The first is that adoption is early, as only 23% of U.S. establishments have built AI into regular workflows, and most AI-enabled productivity gains remain in the future.

The second is the economy. Given that the economy and labor market have remained resilient in 2026, companies may have had less incentive to focus on aggressively cutting labor costs.

For now, the layoff gauge reads low. AI-attributed layoffs in Challenger's tracking have pulled back sharply from their spring peak, and news coverage of AI-related job losses has fallen back to levels from the start of 2026. "While these patterns suggest that a modest labor market headwind is emerging, the near-term headwinds remain smaller than feared earlier this year," Briggs wrote.

Does Displaced Mean Unemployed for Good?

Not in Goldman's view, since Briggs says the displacement is temporary. In the October analysis, Briggs measures gross job losses rather than unemployment, because "some workers that are displaced may find jobs relatively quickly," Briggs wrote, "while others may decide to exit the labor force."

Goldman's long-run view is that the losses fade. As long as the job losses are spread out enough over a 10-year period, Briggs expects the unemployment rate increase in any given year would be less than one percentage point. His reasoning is that around 85% of job growth has been driven by the technological creation of new positions over the last 80 years.

The US job market already churns at scale, with around 30 million jobs being created every year as millions of others are destroyed. In a world where AI is enabling innovation, even a five percent acceleration in the pace at which new jobs are being created would be more than enough to reabsorb workers displaced by automation, in Briggs's estimate.

We do expect that ultimately there are going to be more than enough jobs created to reabsorb workers back into the labor market, Joseph Briggs said on the podcast, according to the Goldman Sachs Exchanges transcript.

The risk Goldman itself names is speed. The job losses could be concentrated over a shorter period, resulting in a notably larger rise in the unemployment rate, and the June report also lists the possibility that AI is fundamentally a more labor-replacing technology than past ones.

What Happens to Workers After a Technology Job Loss, and Does Retraining Help?

Workers displaced by earlier technology take about one month longer to find a new job and incur real earnings losses of over 3% upon reemployment, compared with negligible losses for workers displaced from more stable occupations. The main cause is occupational downgrading, as technology-displaced workers are more likely to move into roles with higher routine content.

Recessions make that worse, which links back to the new October note. Workers displaced by technology during downturns spend about three more weeks unemployed and experience a roughly 5pp higher risk of future unemployment and labor force exit than other displaced workers.

Retraining can also help cushion these impacts. Workers who join retraining programs within three years of a job loss achieve ~2pp stronger cumulative real wage growth over the next decade, and their probability of future unemployment declines by ~10pp. Retrained workers tend to move into roles with higher abstract content and greater complementarity with technology, which Goldman says likely reduces their exposure to further automation.

Who you are also matters. Younger, college-educated and urban workers who lose jobs to technology tend to experience roughly half the cumulative earnings losses of other technology-displaced workers, and workers with shorter job tenure also tend to fare better.

Where Is AI Creating Jobs Instead?

The build-out behind AI is hiring. As of Goldman's March update, construction jobs exposed to the data center build-out had increased by 216,000 since 2022, and relative to trend, the hiring for HVAC contractors, electrical contractors, and other workers to build data centers has risen.

Power demand adds more. Evan Tylenda of GS SUSTAIN estimates that roughly 500,000 net new jobs will need to be filled to satisfy the growing demand for power by 2030 in the US alone. The market will also need construction workers, engineers, electricians, and lineworkers.

Briggs also lists three kinds of new jobs the AI revolution can create. They are roles for workers with AI knowledge and associated skills, the demand mapped in what tech workers need to learn as AI skills take over job postings, new specialized occupations in fields like healthcare, and jobs driven by rising incomes, such as the 1 million workers employed in pet care, nail salons, educational support/tutoring, and athletic coaching cited in Goldman's March outlook.

What Help Can a US Worker Get After an AI-Related Layoff?

The WIOA Dislocated Worker program is designed to help workers get back to work as quickly as possible and overcome barriers to employment. When people become dislocated workers as a result of job loss, mass layoffs, global trade dynamics, or transitions in economic sectors, the program provides services to help them re-enter the workforce.

Services run through a national network of American Job Centers, and one service the program funds is the Rapid Response service. You can find a local center through CareerOneStop or ETA's toll-free help line at 1-877-US-2JOBS.

Services are designed to meet local needs and may vary from state to state, and some carry eligibility rules, so your local center is the place to confirm what you qualify for. The WIOA Adult program provides job search assistance and training opportunities.

What This Means for Workers, and What Should You Do Next?

The October note changes the question from how many jobs to when. If job losses cluster in the next downturn, the time to prepare is while the market is still resilient, not after layoffs begin. That is our reading of the research, and it points to a few concrete steps.

• Check how routine your work is. History shows routine roles take the hit in downturns, as the routine share declined sharply during downturns. If most of your day is repeatable processing, start shifting toward tasks that need judgment.
• Move toward work that complements AI. Retrained workers tend to move into roles with higher abstract content and greater complementarity with technology, which Goldman says likely explains their stronger outcomes, so learn the AI tools in your field rather than avoiding them.
• Watch the layoff signal Goldman watches. Briggs expects AI-related layoffs to provide a good real-time guide, so a rising count of AI-linked cuts in your industry is a reason to update your resume early.
• Consider the trades behind AI. As of Goldman's March update, construction jobs exposed to the data center build-out had increased by 216,000 since 2022, and the market will also require skilled technical work, of the kind provided by construction workers, engineers, electricians, and lineworkers.
• Know your safety net before you need it. The Dislocated Worker program works through American Job Centers, so look up your nearest center before a layoff rather than after one.

Entry-level workers are the group where AI displacement could be stronger, though younger workers have also historically adjusted more quickly to technology-induced shifts in labor demand.

Related Articles

• Australia Says AI Has Not Triggered Mass Layoffs Yet and the Data Backs It Up
• The US Labor Market Is Splitting in Two as AI Cuts Office Jobs While Construction Hires
• The AI Boom Needs Data Centers, But Not Enough Workers to Build Them
• How to Job Search With AI After a Layoff
• How One Nonprofit Is Reskilling American Workers for the AI Era
• Amazon, OpenAI and Anthropic Join Forces to Retrain US Workers for AI
• What the Messy AI Jobs Research Actually Tells Job Seekers
• Companies Are Quietly Rehiring the Workers They Replaced With AI
• How to Stop AI From Becoming the Enemy of Younger Workers

People Also Asked

Q: How many US jobs does Goldman Sachs say AI could displace?

A: Briggs said nine percent of US workers being displaced would correspond to 15 million workers over a 10-year period, and the October note adds that slower hiring could weigh on employment by another 3%.

Q: Why would a recession make AI job losses worse?

A: Companies have historically tended to automate work and shed labor costs when the economy starts to slow, so Briggs wrote that a cyclical slowdown would likely accelerate AI-related automation and job losses.

Q: Will AI cause permanent mass unemployment?

A: Goldman does not expect that, because it sees new job creation absorbing displaced workers over time and, as long as the losses are spread out enough over a 10-year period, expects the unemployment rate increase in any given year would be less than one percentage point.

Future-proof your career? Metaintro brings together 50M jobs in one place. Create a free profile and start your search today.