300mn jobs exposed: Inside Goldman Sachs’ AI labour market warning

The change is not yet enough to suggest that AI is causing a sweeping collapse in employment. Instead, the pressure is concentrated in particular industries and occupations and increasingly among people at the beginning of their careers. AI may not simply eliminate jobs. It could change which workers companies need, what skills they value and where new employment is created. However, the first signs are appearing in sectors where AI can already perform a meaningful share of human tasks.

The warning signs are showing up in AI-exposed industries

Information and communication services are among the industries most exposed to AI. Employment growth in the sector has slowed across almost all major developed economies since 2022. There is, however, an important caveat. Outside the US, employment in these industries remains close to or above its long-run trend. This shows the presence of AI does not automatically translate into widespread job losses. Adoption, demand and the broader state of each economy still matter.

The picture becomes more striking when Goldman looks at individual industries with particularly high exposure to AI. Employment in call centers, software publishing, management consulting and advertising has fallen significantly below its historical trend across developed markets.

Call centers stand out

Employment in the industry is now 39% below trend in the US, 33% below trend in Canada and 27% below trend in Germany. Those numbers shows what AI-driven automation can look like when technology capable of performing human tasks is already available and being deployed. The pattern, Goldman says, indicates that AI-related employment pressures are already visible in industries where automation tools have become practical. But it is still too early to say that AI has transformed the labour market as a whole. Goldman’s conclusion is more measured, AI-related hiring pressures are clearly visible globally, but they remain concentrated in a relatively narrow group of industries and workers.

The first rung of the career ladder is getting harder to climb

The most important finding is not about which industries are affected, but which workers are most exposed. Goldman analysed employment growth across more than 800 occupations and found that AI-related headwinds have been strongest among entry-level workers. There is also an additional, though smaller, negative effect in occupations considered particularly vulnerable to displacement by AI. The difference between the workforce and new entrants is significant. Across the labour market, a 10% occupational exposure to AI was associated with only a 0.1 percentage point drag on annual headcount growth in France, Canada and the US. For entry-level workers, the effect was considerably larger. The estimated impact ranged from more than 0.6 percentage point in Australia to over 0.2 percentage point in the US

This creates a difficult question for young workers entering industries that traditionally gave a path into professional careers. If AI can perform some of the routine tasks that once helped junior employees build experience, that is writing, research, analysis, customer support, coding or basic design, for example, companies may have less reason to hire as many people at the bottom of the career ladder. The issue is therefore bigger than outright job losses. It is also about whether there will be enough entry-level opportunities for workers to acquire the experience needed to move into higher-skilled roles.

Goldman has been warning about this transition for years

The latest findings build on an earlier Goldman Sachs assessment of how AI could reshape employment. Joseph Briggs, who co-leads the global economics team at Goldman Sachs Research, said the technology was already having an impact in specific pockets of the US economy. “You can see AI’s impact in the tech sector, where the employment share as a proportion of the whole economy has gone below the long term trend,” he says in the report.

At the time, however, the US labour market did not yet show significant evidence of an AI-led shift in the overall employment mix.This could change as adoption moves from experimentation to widespread deployment. In Briggs’ base case, it could take around 10 years for companies to adopt AI on a broad scale. During that transition, Goldman estimates that 6%-7% of workers could be displaced. The speed of that process will be critical. If companies gradually introduce AI over a decade, Goldman expects the unemployment rate to rise by around 0.6 percentage point. A much faster transition would have a significantly larger economic impact. That is why the question is not simply how many jobs AI can automate. It is how quickly businesses decide to automate them.

Around 300 million jobs are exposed globally

The potential scale is enormous. Goldman estimates that around 300 million jobs globally are exposed to AI automation. In the U.S., AI could potentially automate tasks that currently account for 25% of all work hours. However, exposure does not mean elimination. A job can contain tasks that AI can perform without the entire occupation disappearing. In many cases, workers may use AI to complete some parts of their jobs faster while continuing to perform tasks that require judgment, communication, creativity or physical presence. Even so, some occupations are already showing signs of displacement. Management consultants, call center workers and graphic designers are among knowledge and creative workers who have experienced some labour displacement linked to AI. The next phase could be different. As AI becomes cheaper, more capable and easier for companies to deploy, the technology could move beyond isolated use cases and begin changing hiring decisions across a much wider range of businesses.

The real problem could be a skills mismatch

There is another side to the AI labour-market story. The workers displaced from one part of the economy may not necessarily have the skills needed for the jobs being created elsewhere. Evan Tylenda, an analyst with GS SUSTAIN, says in the report that workers pushed out of knowledge industries by AI may be less suited to some of the occupations where demand is expected to be strongest. Some of those jobs are relatively low-skilled and lower-paid. Fast-food workers, cleaners and home healthcare workers are examples of occupations that could continue to see demand.

At the other end of the spectrum, the economy is expected to need more skilled technical workers, including construction workers, engineers, electricians and lineworkers.This creates a potentially uncomfortable transition. A worker whose job is disrupted by AI cannot necessarily move overnight into an electrical, engineering or construction role. The problem, therefore, may not be simply unemployment. It could also be the time and investment required to move people from declining occupations into expanding ones.

AI boom is already creating jobs around data centers

The infrastructure being built to power AI provides one early example of this shift. Goldman says hiring for HVAC contractors, electrical contractors and other workers involved in data-center construction has risen relative to its historical trend. Construction jobs exposed to the data-center build-out have increased by 216,000 since 2022. “And we expect that data center investment will continue to grow,” Briggs says in the report. AI requires enormous computing infrastructure, and that infrastructure requires electricity, cooling systems, buildings and networks. In the US alone, roughly 500,000 net new jobs will need to be filled to meet growing demand for power by 2030, Tylenda says,which means the AI economy could simultaneously reduce demand for certain white-collar tasks while creating demand for workers who build and operate the physical infrastructure behind the technology.

AI may create jobs that do not exist yet

The future labour market will not be defined only by jobs disappearing. Briggs expects AI to generate three broad categories of new employment. The first is jobs that require AI knowledge and related skills. As companies adopt the technology, they will need people who can build, manage, implement and work alongside AI systems.

The second is entirely new specialist occupations. Healthcare provides one example. Technological advances in the past have created opportunities for greater specialization in medicine and other fields. AI could produce similar specialization across a range of industries.

The third category is less obvious. AI could indirectly create jobs by increasing productivity, incomes, demand and the availability of workers. Some occupations that are now common did not necessarily emerge because technology suddenly made them possible. They became viable because rising incomes and changing consumer preferences created enough demand.

Goldman points to pet care, nail salons, educational support and tutoring, and athletic coaching. Around 1 million workers are employed in such occupations today, many of which emerged over the past 30 years. That suggests the jobs created by the AI economy may not all have “AI” in their titles.

Developed economies are adopting AI at different speeds. The scale of the labor-market disruption will also depend on how quickly countries adopt the technology. Goldman combined 11 surveys measuring AI adoption across countries and found that major developed economies currently have adoption rates of roughly 15%-20%.

France, the US, the Netherlands and the UK are among the leaders. Italy, Japan and New Zealand are closer to the lower end among developed economies. Major emerging markets have estimated adoption rates of around 10%-15%. Those differences could become increasingly important. A country where businesses rapidly integrate AI into everyday operations may see productivity gains earlier but it could also experience employment disruption sooner. Countries that adopt more slowly may have more time for workers and institutions to adjust, although they could also miss out on some of the productivity gains.

The US is entering an important year for jobs

The AI question comes at a particularly sensitive moment for the US labour market. Goldman has been watching the US closely as job growth slowed in the second half of 2025. Briggs identifies two major forces behind that slowdown. The first was uncertainty around tariffs, which made employers less confident about hiring. The second was a dramatic slowdown in immigration after a rapid pace of immigration in 2023 and 2024. Goldman expects net immigration to slow further this year.

The bank nevertheless expects the labor market to stabilize, helped partly by fiscal tailwinds and other factors that are forecast to drive a sharp acceleration in GDP growth in the first half of 2026. Unemployment is estimated to rise modestly to 4.5% this year, from 4.3% in January. However AI adds another variable to the outlook. “The big story in 2026 in labor will be AI,” Briggs says. If companies accelerate AI-related job cuts and bring those losses forward, the labor market could underperform Goldman’s current expectations. This could also have consequences for monetary policy. A weaker labour market could increase the likelihood of the Federal Reserve cutting interest rates.

The biggest AI story may be about who gets hired next

For workers in their 20s and 30s entering knowledge and content-creation industries, the transition could be especially important. These are precisely the kinds of jobs where AI is becoming capable of handling increasingly sophisticated tasks. But Goldman is not predicting a predetermined outcome. “But this is not a foregone conclusion,” Briggs says.

The technology could replace some tasks, augment others and create entirely new categories of work. Companies could also use productivity gains to expand rather than simply reduce their workforce. The uncertainty is therefore substantial. “There’s a lot of uncertainty about how this will play out. It’s important to watch the labor market data in real time to monitor that,” he adds.