Which Jobs Are Actually Disappearing to AI Right Now

Goldman Sachs Research estimates that 300 million jobs globally are exposed to AI automation, and the first wave of displacement is already visible in hiring data. AI has reduced monthly payroll growth by roughly 16,000 jobs in the United States over the past year, and the unemployment rate has ticked up by 0.1 percentage point as a result (Goldman Sachs, 2026). The shift is narrow today, concentrated in a handful of sectors, but the trajectory is clear: AI is no longer a future risk. It is a present headcount.

The question that matters for workers is not whether AI will change the job market. It already has. The question is which roles are being substituted, which are being augmented, and where the line falls between a job that AI makes more productive and a job that AI makes redundant. The data from 2025 and 2026 offers the first real answers, and they are not evenly distributed across age groups, industries, or seniority levels.

A Stanford/NBER study tracking millions of U.S. workers through June 2026 found no evidence of widespread, economy-wide job displacement. But the study identified a sharp divergence: employment of young workers ages 22 to 25 in AI-exposed occupations stands 19% below where it would have been if it had kept pace with their less-exposed peers (Brynjolfsson et al., 2026). That gap has widened steadily since the researchers first documented it in August 2025.

• 300 million jobs globally face AI automation exposure (Goldman Sachs, 2026).
• 16,000 fewer monthly payroll jobs attributed to AI substitution in the past year.
• 6-7% of U.S. workers expected to be displaced over a 10-year adoption cycle.
• 22-25-year-olds in AI-exposed roles: 19% below expected employment levels.

Which industries are seeing the fastest AI-driven job losses?

Goldman Sachs found that employment in call centers, software publishing, management consulting, and advertising services has fallen sharply below historical trends across developed markets. In the United States, call center employment is now 39% below trend (Goldman Sachs/CNBC, 2026). Canada sits at 33% below trend, and Germany at 27%. The pattern is not subtle: industries where AI tools capable of automating tasks are already deployed are seeing real hiring declines.

The Goldman Sachs analysis, which examined employment growth across more than 800 occupations, found that information and communication services, among the industries most exposed to AI, have slowed across nearly all major developed economies since 2022. The relationship is particularly pronounced in Germany, Australia, and the United States (CNBC, 2026). Employment growth in graphic design, office administration, and telephone call centers has also fallen below trend amid reports of reduced labor demand.

Are young workers bearing the brunt of AI displacement?

The short answer is yes. The Stanford study found that employment of 22-25-year-olds in the two most AI-exposed quintiles fell about 11% between November 2022 and June 2026, while employment for the same age group in the least exposed occupations grew about 10% — a divergence of 21 percentage points (Brynjolfsson et al., 2026). This is not because young workers are being fired. It is because companies are hiring fewer of them in the first place.

Across the broader labor market, a 10% occupational exposure to AI was associated with only a 0.1 percentage point drag on annual headcount growth for workers generally. But for entry-level workers, the impact ranged from more than 0.6 percentage point in Australia to over 0.2 percentage point in the United States (Goldman Sachs/CNBC, 2026). The pattern is consistent across countries: AI is displacing entry-level hiring before it displaces experienced workers.

What jobs are at the highest substitution risk?

Goldman Sachs economist Elsie Peng found that occupations with similar AI exposure can differ dramatically in whether AI substitutes for or augments human labor. Telephone operators, insurance claims clerks, and bill collectors face the highest substitution risk. By contrast, education workers, judges, and construction managers show the highest AI augmentation potential — roles where AI tools enhance human judgment rather than replace it (Goldman Sachs, 2026).

MIT's Daron Acemoglu, a Nobel Prize-winning economist, puts the current most vulnerable tasks in two categories: cognitive and routine. These are tasks that involve repetitive processes with limited innovation, creativity, or social judgment. Customer service representatives and back-office workers performing data entry, claims processing, and document review fall squarely in this zone. Acemoglu estimates about 8 to 9 million U.S. workers in these roles face near-term displacement pressure (Goldman Sachs, 2026).

Is AI creating jobs at the same time it destroys them?

Goldman Sachs estimates that AI augmentation has increased monthly payroll growth by about 9,000 jobs in the past year, partially offsetting the 16,000-job drag from substitution. The net effect is a modest negative — roughly 7,000 fewer jobs per month than would otherwise exist. But the gross numbers are enormous: 30 million jobs are created and 29 million destroyed every year in the U.S. economy as a baseline. A 5% acceleration in the pace of new job creation would be enough to reabsorb AI-displaced workers, if it materializes (Goldman Sachs, 2026).

The jobs being created tend to cluster in areas adjacent to AI infrastructure. Construction jobs tied to data center build-outs have increased by 216,000 since 2022. Goldman Sachs projects roughly 500,000 net new jobs will need to be filled in data center construction, electrical contracting, and HVAC work. The catch is that these jobs require different skills than the ones being displaced, creating a transition gap that training programs have not yet closed.

How long will the AI job transition actually take?

Joseph Briggs, Goldman Sachs's co-head of global economics, estimates the AI transition will unfold over roughly 10 years. Under that timeline, the peak impact on unemployment in any single year would be less than 0.5 percentage point. But Briggs warns that if adoption is more frontloaded — meaning companies automate faster than expected — the economic impact could be significantly larger. The 10-year timeline is an assumption, not a guarantee.

Acemoglu offers a more cautious outlook. He expects AI to have a small net negative impact on labor over the next five years — far less than technologists predict — but warns that job losses could be larger over 10 to 15 years if investment continues to focus on replacing rather than complementing workers. The wildcard, he says, is the integration of AI and robotics, which could open up the 50% of U.S. economic work that involves physical tasks.

Which jobs are safest from AI displacement?

Goldman Sachs's analysis identifies air traffic controllers, chief executives, radiologists, pharmacists, residential advisors, photographers, and members of the clergy as the occupations least at risk of AI displacement. The common thread is physical presence, high-stakes judgment, and complex social interaction — tasks that AI currently cannot replicate at the cost and reliability needed for production deployment.

The distinction that matters is not just what a job involves, but whether AI automates the most expert or the least expert parts of it. If AI automates the least expert tasks, employment may fall while wages rise as work becomes more specialized. If AI automates the most expert tasks, employment rises while wages fall as more workers can do the job. Neil Thompson, Director of the FutureTech research project at MIT, says this framework suggests the impact will cut across the occupation and wage distribution rather than hitting one worker group in particular.

What should workers do right now?

The data points toward a practical strategy. Workers in cognitive-routine roles — back-office, claims processing, basic customer service, data entry — should be upskilling toward roles that AI complements rather than replaces. That means moving toward work that requires judgment, creativity, physical presence, or complex social interaction. Goldman Sachs finds that college graduates have historically adjusted more nimbly than other workers to technological disruption, which suggests adaptability matters more than any specific technical skill.

• Map your tasks: which are routine and cognitive vs. judgment-heavy and social?
• Upskill toward AI-augmented roles, not AI-substituted ones.
• Track industry hiring trends using BLS and your company's own postings.
• Data center, construction, and healthcare trades are growing fast.

Bottom line: AI is not causing mass unemployment yet, but the data shows real displacement in specific sectors — call centers, consulting, software publishing, and entry-level knowledge work. The transition is narrow enough to be manageable today but wide enough to demand action from workers and employers. The biggest risk is assuming the displacement will stay contained.

• Why AI agents are hitting the payroll
• AI hiring is changing how humans get interviewed
• The nuclear miscalculation risk of AI transparency
• Goldman Sachs: How Will AI Affect the US Labor Market (2026)
• Stanford/NBER: Canaries in the Coal Mine (2026)
• Goldman Sachs: The Jobs AI Is Likely to Boost (2026)
• Data center politics and the future of tech jobs
• The deepseek race to zero and model costs