Goldman Sachs economist Joseph Briggs put a number on something labor economists have argued about all year. In a research note dated October 22, 2026, he wrote that recent labor market data show some signs of AI-related hiring headwinds. Employment growth in the industries most exposed to AI ran roughly 10,000 jobs below its long-term trend over the previous three months.
That figure is small against a US labor force of well over 100 million people. It still matters, because it is the first time a major Wall Street research desk has tied a measurable hiring gap to AI exposure while arguing the bigger shock has not arrived yet. The same note estimates that up to 9% of workers could be displaced during the AI transition. Anyone tracking how AI is already cutting coding time will recognize the pattern: productivity gains show up first, and the labor effects follow with a lag.
This analysis walks through what the Goldman Sachs AI hiring forecast actually says, which claims the data supports, and which viral numbers it does not. It also covers the competing views, the history of automation shocks, and what to watch before the next jobs report.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Goldman Sachs Actually Reported on October 22
Briggs, a senior global economist at Goldman Sachs Research, wrote a note that mixed reassurance with warning. His opening framing was measured: âBut recent labor market data show some signs of AI-related hiring headwinds.â The word “some” carries weight. Goldman is not claiming a collapse in hiring. It is claiming that AI-exposed industries have begun to underperform their own historical trend.
The note then moves to a forecast. Briggs wrote that under the bank’s baseline forecast for a 15% productivity increase from AI, around 9% of workers could be temporarily displaced during the AI transition. Yahoo Finance framed that as one in 11 workers, and added that the layoffs are likely to wait for a recession.
The table below collects the figures Goldman’s note and the press coverage of it agree on. Everything in it traces to the note as reported on October 22 and 23, 2026.
| Metric | Figure | What it means |
|---|---|---|
| Note date | October 22, 2026 | Published by Goldman Sachs Research |
| Hiring shortfall in AI-exposed industries | About 10,000 jobs below trend | Measured over the preceding three months |
| Productivity assumption | 15% AI-driven gain | US and other developed economies |
| Workers displaced in transition | Up to 9% | Roughly one in 11 workers |
| Added drag from slower hiring | Another 3% of employment | On top of displacement |
| Occupations with higher separation rates | 19 of 25 | Average rise of 3.8 percentage points |
| Occupations with lower hiring rates | 13 | Average drop of 0.7 percentage points |
The 10,000-Job Shortfall, Explained
The headline number measures a gap, not a loss. Goldman compared employment growth in the most AI-exposed industries with where long-term trends say it should have landed. The result was about 10,000 jobs short across three months. Those workers were never hired in the first place, which is a different thing from being laid off.
Spread evenly, a 10,000-job gap over three months works out to roughly 3,300 a month. That is simple arithmetic and not a figure from the note, so treat it as a sense of scale only. Against monthly payroll swings that often run into six figures, it sits well inside normal noise, which explains why Briggs hedged his language.
The distinction matters for readers who saw alarming social posts this week. A hiring slowdown in specific industries is a leading indicator. A wave of AI-driven layoffs would be a coincident or lagging one. Goldman’s data shows the first and not the second.
AI Construction Hiring Masks the Gap
One detail keeps the aggregate number tame. Increased hiring in AI-related construction largely offset the shortfall. Data centers, power connections and the buildings around them need electricians, steelworkers, and project crews, and those jobs have been growing as AI spending climbs.
That creates an odd split in the labor market. Knowledge-work categories show softer hiring, while the physical build-out that powers AI absorbs workers. Our earlier look at the AI infrastructure race and its nine model launches in five weeks showed how fast capital is flowing into this layer. The construction boom is real, but it is also narrow and tied to a single spending cycle.
If AI capital spending slows, the offset disappears. The shortfall in exposed industries would then show up in headline employment numbers instead of hiding behind concrete and cable. That dependency is the quiet risk inside an otherwise calm-looking statistic.
Service Industries Show the Clearest AI Unemployment Link
The most statistically solid finding in the note concerns unemployment, not hiring. Goldman found a statistically significant positive relationship between AI adoption and unemployment across service industries. In plain terms, service sectors that adopted more AI saw higher unemployment.
The link between AI adoption and employment growth was weaker, and it pointed in the negative direction. That gap between the two relationships is worth pausing on. Unemployment captures people who lose a job and keep looking. Employment growth also reflects new entrants, retirements and workers who leave the labor force. A cleaner link to unemployment suggests the effect is landing on workers who are displaced rather than on overall headcount.
Goldman’s published materials on the topic, including its piece on how AI will affect the labor market, have stressed this theme for months. Briggs asked in that discussion whether new job creation would pick up quickly enough to offset any near-term headwinds: âThe bigger question is: does that pace of new job creation pick up fast enough to offset any term headwinds?â The October note gives that question its first hard data point.
Separations Up in 19 of 25 Occupations
The occupation-level numbers show where the pressure sits. Separation rates, which count people leaving jobs for any reason, rose in 19 of 25 occupations. The average increase was 3.8 percentage points. Hiring rates fell in 13 occupations, with an average drop of 0.7 percentage points.
Compare those two averages and a pattern emerges. The larger move came on the exits side, not the entrances side. Workers are leaving jobs faster than firms are cutting hiring. That fits a labor market where employers still hire but where people are moving, being moved, or being pushed out of roles that automation touches.
A caution applies. The two averages cover different sets of occupations (19 versus 13), so they are not a like-for-like comparison. Separations also include voluntary quits, which can rise when workers feel confident, so a higher rate does not automatically signal distress. Goldman’s reading is that AI is a contributing factor, not the sole driver.
The 9% Displacement Scenario and the 3% Hiring Drag
The forecast portion of the note deserves careful reading. Goldman’s baseline assumes AI lifts productivity by 15% in the United States and other developed economies. Under that scenario, up to 9% of workers could be displaced during the transition. Slower hiring could then weigh on employment by another 3%.
Briggs used the word “temporarily” in describing the displaced group: âUnder our baseline forecast for a 15% productivity increase from AI, around 9% of workers could be temporarily displaced during the AI transition.â That qualifier is the difference between a one-time reshuffling and a permanent loss of work. History suggests displaced workers do find new roles, but the path can be slow and uneven.
The scenario also depends on an assumption that may or may not hold. A 15% productivity gain is a forecast, not an observation. If AI delivers less, displacement shrinks. If it delivers more or faster, the transition gets harder. Goldman’s number is a planning anchor, and readers should treat it that way.
Why Economists Expect the Damage to Wait for a Recession
The most useful idea in the note may be about timing. Briggs reviewed decades of data on technology and employment and found that automation-related disruption stays limited when the economy is healthy but accelerates during slowdowns. Companies tend to automate work and trim labor costs when growth fades. Fortune reported that concerns about imminent AI displacement had cooled since peaking in the spring.
Briggs put the cyclical risk bluntly: âAI-related labor disruptions could well amplify the broader business cycle.â The line comes from Fortune’s coverage of the Goldman forecast. Read it as a statement about feedback loops. A downturn pushes firms to automate, automation removes jobs, and fewer jobs deepen the downturn.
This framing changes how to read a calm labor market. Low layoffs today do not prove AI is harmless. They may only show that firms have no pressing reason to restructure. With the Federal Reserve meeting on October 28, as covered in our piece on markets waiting on the Fed and the midterms, any hint of a slowdown would test Goldman’s thesis quickly.
Pantheon’s 10,000-a-Month Claim Versus Goldman’s Evidence
A second number has circulated alongside Goldman’s. Fortune published a piece on October 23 under a headline saying AI is killing 10,000-plus jobs per month and that 9% of jobs are at risk, citing analysts. Available reporting attributes the monthly figure to Samuel Tombs and Oliver Allen of Pantheon Macroeconomics.
Here the evidence gets thin. Goldman’s note confirms a shortfall of about 10,000 jobs relative to trend, measured over three months. It does not confirm a loss rate of 10,000 jobs every month. The two statements sound alike and describe different things: one is a cumulative gap against a trendline, the other is a flow of jobs destroyed each month.
Readers should treat the monthly claim as unconfirmed. Nothing in the Goldman data supports it, and we could not verify the Pantheon methodology. The honest summary is that two analyses both land near 10,000, yet they measure different quantities over different periods. Merging them into a single “10,000 jobs a month” story would overstate what the data shows.
Goldman’s CEO Rejects the Job Apocalypse Framing
Goldman’s research desk and its chief executive are not saying the same thing in the same tone. In a January interview, David Solomon, chairman and CEO of Goldman Sachs, told Fortune: âIâm not in the job apocalypse camp.â He added that technology has been disrupting and destroying jobs for decades while pushing the economy to create new ones, and that âItâs no different this time.â
Solomon also said that if the firm gets its AI integration right, it does not significantly lower the number of people it employs. That is a statement about Goldman’s own workforce, not the economy, and it came nine months before Briggs’s note. The two views are compatible. Solomon describes how a firm can use AI to expand capacity, while Briggs describes aggregate risk if a downturn arrives.
Enterprise adoption is also moving unevenly. Our reporting on shadow AI use among 45% of workers shows that employees are adopting tools faster than official programs, which complicates any attempt to link adoption statistics to hiring.
Competing Readings of the AI Hiring Headwinds
The AI impact on jobs is a contested subject, and the sources covered here do not agree on severity or timing. The table lays out who says what, using only claims drawn from the reporting cited in this article.
| Source | Position | Evidence cited | Status |
|---|---|---|---|
| Joseph Briggs, Goldman Sachs Research | Early hiring headwinds, bigger risk in a downturn | 10,000-job shortfall, 19 of 25 occupations with higher separations | Confirmed in note |
| Samuel Tombs and Oliver Allen, Pantheon Macroeconomics | AI is destroying 10,000-plus jobs per month | Attributed via Fortune’s Oct. 23 headline | Unconfirmed against Goldman data |
| David Solomon, Goldman Sachs CEO | No job apocalypse, as in past tech shifts | Historical pattern of job creation | Stated in January interview |
| Members of Congress | Prepare aid for workers if mass AI layoffs come | Reported by The New York Times on Oct. 23 | Early policy discussion |
The Congress item deserves a note. The New York Times reported on October 23 that some lawmakers fearing mass AI layoffs are weighing ways to help affected workers. Policy attention usually trails the data, so the appearance of that discussion suggests the political class is reading the same early signals as Goldman.
Frontier Model Releases Raise the Stakes
Hiring headwinds are showing up while model capability and price competition keep intensifying. Recent weeks brought a crowded run of frontier releases. Our coverage of Gemini 4 Argon, with its 1M-token context and $2 input price and of Anthropic’s Claude Opus 5.5 agent shows how quickly cheaper, longer-context models are reaching paying customers.
Better models widen the set of tasks that software can handle at acceptable cost. That does not map directly onto job losses, because firms often use new capability to raise output rather than cut staff. Still, a lower price per token improves the economics of automating routine knowledge work, especially in service industries where Goldman found the clearest unemployment link.
Reports on the note also point to limited adoption so far. A technology that many firms have not yet built into regular workflows cannot have produced its full labor effect. As integration deepens, the same models that look like experiments today will sit inside payroll-relevant processes.
Market Impact: What Investors and Employers Watch Next
Markets will read the Goldman note through two lenses. For AI-exposed equities, evidence that firms are slowing hiring because of automation supports the productivity thesis behind current valuations. For consumer-facing sectors, softer service-sector employment threatens household spending. Both readings will feed into the heavy calendar ahead, including the Big Tech earnings run on October 28 and 29, where management teams will talk about AI efficiency and headcount.
For investors
Watch for language on headcount in earnings calls. Flat or falling staff numbers alongside rising revenue would confirm the productivity channel Goldman describes. Our AI and machine learning coverage tracks these model and spending announcements as they land.
For employers and workers
Employers face a timing choice. Automating during a strong economy carries little political cost, while doing it in a slowdown draws scrutiny. Workers in service roles and routine occupations should treat the separation data as a prompt to build skills that sit beside AI tools instead of in competition with them.
Historical Context: From Jobless Recovery to Jobless Growth
Goldman has made this timing argument before. Fortune noted that in October 2025, Goldman US economists David Mericle and Pierfrancesco Mei called jobless growth the likely new normal and warned that the full labor-market consequences of AI might not be visible until a recession. Briggs reached a similar conclusion a year later from a different team.
The historical chart in the new note tracks the share of Americans in routine jobs, such as clerks, secretaries, machine operators and mechanics. According to Fortune’s account, that share stays stable in expansions and drops sharply in downturns, with step-downs around the recessions of the early 1980s, 2001, 2008 and 2020. It does not climb back to its prior level afterward.
That record is why the current calm is hard to interpret. The same data that supports “limited disruption so far” also supports “disruption concentrated in the next recession.” The sample is small, with only a handful of recessions, so the pattern is suggestive rather than proven.
Five Predictions for AI Hiring Headwinds
The following are our own forecasts based on the sources above. They are judgments, not reported facts, and any could prove wrong.
| # | Prediction | Indicator to watch | Risk to the call |
|---|---|---|---|
| 1 | The shortfall in AI-exposed industries persists through year-end | Monthly payroll data for exposed sectors | Strong construction hiring could mask it again |
| 2 | The unverified “10,000 a month” claim keeps spreading | Repeats in headlines and social posts | Pantheon or Goldman could clarify methodology |
| 3 | Earnings calls feature more headcount-versus-revenue talk | Oct. 28-29 Big Tech results | Firms may avoid the subject for political reasons |
| 4 | Congress moves from discussion to draft worker-aid proposals | Bill introductions after the midterms | Election outcomes could stall the effort |
| 5 | A slowdown would raise AI-attributed layoffs sharply | Layoff trackers during any downturn | A soft landing would delay the test |
The common thread is that the data so far is early and mixed. A single three-month window cannot settle the argument. The next several jobs reports, and how the economy behaves around them, will say far more than any one forecast.
Frequently Asked Questions
What did Goldman Sachs say about AI hiring headwinds?
In an October 22, 2026 note, Goldman economist Joseph Briggs said recent labor market data show some signs of AI-related hiring headwinds. Employment growth in the industries most exposed to AI was about 10,000 jobs below its long-term trend over the preceding three months.
How many workers could AI displace, according to Goldman?
Under Goldman’s forecast that AI raises productivity by 15% in the United States and other developed economies, up to 9% of workers could be displaced during the transition. Slower hiring could affect employment by another 3%.
Is AI destroying 10,000 jobs a month?
That claim is unconfirmed. Reporting attributes it to Pantheon Macroeconomics analysts, but Goldman’s note supports only a 10,000-job shortfall relative to trend over three months. Those are different measures.
Why did construction hiring matter in the note?
Increased hiring in AI-related construction largely offset the shortfall in exposed industries. That means the aggregate number looked calmer than the underlying weakness in AI-exposed service work.
Which occupations saw the biggest changes?
Separation rates rose in 19 of 25 occupations, by an average of 3.8 percentage points. Hiring rates fell in 13 occupations, by an average of 0.7 percentage points. Goldman also found a significant positive link between AI adoption and unemployment in service industries.
Does Goldman think mass AI layoffs are coming now?
No. Briggs argued that automation-related disruption tends to stay limited in healthy economies and speed up during downturns. Fortune reported that worries about imminent displacement had cooled since the spring.
What is the main takeaway for workers?
The data points to early pressure on hiring in AI-exposed service roles, not a collapse. The takeaway: the larger risk is tied to economic cycles, so skills that complement AI tools and a watch on the broader economy matter more than any single headline number.
