Goldman Sachs finds that AI-sensitive industries are hiring more slowly across developed economies, but the disruption remains concentrated in areas such as call centers, software and advertising.
MARKET INSIDER — Artificial intelligence is beginning to weaken employment growth in some developed economies, with the clearest pressure falling on entry-level workers and industries where existing tools can already automate routine cognitive tasks, according to new Goldman Sachs research.
The effects remain narrow rather than economy-wide. Call centers, software publishing, management consulting and advertising have fallen furthest below their historical employment trends, while the overall relationship between AI exposure and job growth remains modest.
The findings suggest AI is changing how companies hire before it produces widespread job losses—particularly by reducing demand for junior employees performing standardized work.
Key Highlights
- AI-exposed industries have generally recorded weaker job-opening growth since late 2022, especially in Germany, Australia and the United States.
- Call-center employment is 39% below trend in the U.S., 33% below trend in Canada and 27% below trend in Germany.
- The employment effect remains small across the overall workforce but is several times stronger for entry-level workers.
AI pressure is visible—but still concentrated
Goldman examined employment and job-opening data across major developed economies, comparing recent performance with historical trends and industries’ exposure to AI automation.
Information and communication services—among the sectors most exposed to generative AI—have experienced slower employment growth across nearly every developed market since 2022.
Outside the United States, however, employment in these industries generally remains near or above its long-term trend. That suggests AI has slowed hiring more consistently than it has caused widespread net job destruction.
The distinction matters. A company can reduce graduate recruitment, leave vacant positions unfilled or rely on natural employee turnover without announcing layoffs. The resulting labor-market adjustment may therefore appear first in job openings and opportunities for new workers rather than headline unemployment.
Goldman found the relationship between AI exposure and slower job-opening growth was particularly evident in Germany, Australia and the U.S.
Call centers provide the clearest displacement signal
Employment has fallen substantially below historical trends in call centers, software publishing, management consulting and advertising.
Call centers stand out because modern AI systems can already perform a significant portion of their core activities. Voice agents and chatbots can answer common questions, authenticate customers, process simple transactions and route complex cases to human employees.
Goldman estimated that call-center employment was 39% below trend in the U.S., 33% below trend in Canada and 27% below trend in Germany.
These figures measure the gap from estimated historical trends, not necessarily the number of positions eliminated directly by AI. Other factors—including outsourcing, weaker economic growth and previous generations of customer-service automation—may also have contributed.
Nevertheless, the consistency across several countries strengthens the argument that AI is now affecting industries where deployable automation tools are readily available.
In software publishing, AI coding assistants may allow developers to complete routine tasks more quickly. Advertising agencies can automate initial drafts, image production, audience analysis and campaign variations. Consulting firms can use AI for research, document review and presentation preparation.
These tools do not eliminate the need for experienced professionals, but they can reduce the amount of junior labor required to produce the same output.
Entry-level workers face disproportionate pressure
Goldman’s analysis of more than 800 occupations found that AI-related employment headwinds were strongest among entry-level workers.
Across the broader labor market, a 10% increase in occupational AI exposure was associated with only a 0.1-percentage-point drag on annual headcount growth in France, Canada and the United States.
For entry-level workers, however, the estimated effect ranged from more than 0.6 percentage point in Australia to over 0.2 percentage point in the U.S.
Junior positions are particularly exposed because they frequently involve structured work that can be reviewed by more experienced employees: preparing first drafts, summarizing information, conducting basic research, formatting presentations, writing standard code and responding to routine customer requests.
These tasks also traditionally help younger employees acquire the knowledge needed for more senior roles. If companies automate too much of this work, they could weaken the training pipeline that produces experienced workers later.
Businesses may consequently need to redesign entry-level employment rather than simply remove it. Junior employees could spend less time producing routine materials and more time validating AI output, communicating with clients and learning judgment-intensive work.
Correlation does not establish that AI caused every decline
The timing creates an important analytical problem.
The second half of 2022 also marked a wider slowdown in technology hiring following the pandemic-era expansion. Higher interest rates, weaker venture-capital funding and cost-cutting across digital businesses affected many of the same industries considered highly exposed to AI.
Employment falling below trend in software or advertising therefore cannot be attributed entirely to generative AI.
Goldman’s cross-country and occupation-level findings strengthen the evidence of an AI effect, but they do not remove all competing explanations. Differences in economic growth, employment regulation, outsourcing and industry composition can influence each country’s results.
Other research remains more cautious. A 2026 study comparing AI adoption in the United States and Europe found productivity benefits but no clear evidence that recent adoption had yet produced systematic employment changes across either region, according to Brookings Institution
The most defensible conclusion is therefore that AI-related pressure has become measurable in selected occupations—not that it is already driving a broad labor-market contraction.
Adoption is spreading unevenly
Goldman combined 11 surveys to estimate AI adoption across countries. Its composite measure placed adoption at approximately 15% to 20% in major developed economies and between 10% and 15% in large emerging markets.
France, the U.S., the Netherlands and the U.K. ranked among the more advanced adopters. Italy, Japan and New Zealand were closer to the lower end among developed economies.
Adoption estimates vary substantially depending on whether a survey measures occasional individual use, company-wide implementation or AI deployed directly in production. This explains why other studies sometimes report much higher usage rates.
For example, a Federal Reserve Bank of St. Louis survey found that 43% of U.S. workers had used generative AI at work in early 2026, but AI accounted for only about 5.2% of total U.S. working hours.
Using an AI tool and reorganizing a workforce around it are very different stages of adoption. Employment effects are likely to become stronger only when companies redesign processes rather than merely give workers access to chatbots.
What it means for companies and investors
For companies, the earliest financial benefit may come through slower hiring rather than mass layoffs. Reducing recruitment, limiting contractor use and increasing output per employee can improve margins without the restructuring expenses and reputational damage associated with large staff reductions.
Labor-intensive business-services companies face greater disruption. Call-center operators, outsourcing providers, advertising agencies and lower-value software contractors may experience pricing pressure as customers expect AI-related savings.
Companies possessing valuable proprietary data, established customer relationships and employees capable of supervising AI may be better positioned. Their technology can complement human expertise rather than compete primarily by replacing inexpensive labor.
For investors, the critical indicators include revenue per employee, headcount growth, entry-level recruitment, restructuring costs and whether productivity gains improve margins or are passed to customers through lower prices.
Implications for Asia and emerging markets
The immediate employment effect may be smaller in emerging economies because measured AI adoption remains lower. However, countries dependent on outsourced services could eventually face greater structural exposure.
India, the Philippines and other Asian markets have built large employment sectors around call centers, back-office processing, software services and standardized administrative work. These are among the activities most susceptible to AI-enabled productivity gains.
Lower wages provide some protection because automation must deliver sufficient savings to justify implementation costs. That advantage may weaken as AI systems become cheaper and more reliable.
Emerging economies will need to move employees toward work requiring industry knowledge, relationship management, multilingual communication, system supervision and complex judgment.
Goldman’s findings do not indicate that AI is producing economy-wide unemployment. They show something more targeted: companies are beginning to need fewer workers in occupations where routine cognitive tasks can already be automated—and young people trying to enter those professions are feeling the change first.
