Numbers don’t lie. Employment in sectors most exposed to artificial intelligence has slipped while the broader economy added jobs. The trend, once theoretical, now shows up in government statistics and private analysis. But the story isn’t simple replacement. It’s a tale of who gets hurt first and what comes next.
Signs of Trouble in the Data
Customer service representatives saw their ranks shrink by 130,180 positions, a 4.8 percent drop over the year ending in May 2025. That category forms part of a larger group of 18 occupations the Bureau of Labor Statistics flagged years ago as tied to AI advances. Overall, those 18 roles posted a 0.2 percent decline from May 2024 to May 2025. Meanwhile total U.S. employment rose 0.8 percent. Remove medical secretaries and administrative assistants, whose numbers grew and may distort the picture, and the remaining AI-linked jobs fell 1.6 percent. (Gizmodo)
Since ChatGPT launched in late 2022, the pattern sharpened. Total employment climbed 2.5 percent. Yet jobs in the 10 percent of sectors with highest AI exposure, using an index from economists Edward W. Felten, Manav Raj and Robert Seamans, dropped 1 percent. Computer systems design and related services suffered a steeper 5 percent decline. The losses hit young workers hardest. Employment for those under 25 in these areas fell noticeably. Older workers saw no such drop. Dallas Fed economist Tyler Atkinson put it plainly: “this fall in employment for those under 25 is not due to layoffs but to a low job finding rate for young workers. Basically, the job market is getting very tough for new graduates in AI-exposed fields.” (Dallas Fed)
Goldman Sachs economists took the analysis further. They separated substitution, where AI takes over tasks outright, from augmentation, where it boosts human output. Substitution erased about 25,000 jobs per month over the past year. Augmentation added back roughly 9,000. Net result: 16,000 fewer jobs monthly. The pain concentrates on Gen Z. These workers cluster in routine white-collar positions such as data entry, billing and customer support. Those roles match exactly what current AI systems handle well. (Fortune)
And the experience gap widened. In occupations with high substitution risk, the unemployment difference between entry-level workers under 30 and those aged 31 to 50 grew larger than before the pandemic. A one-standard-deviation rise in substitution exposure pushed that entry-level wage gap up by 3.3 percentage points. Young graduates fluent in AI tools still struggle. Their knowledge often stays codified. AI masters that part fast. Tacit knowledge, built through years on the job, proves harder to replicate. Stanford researchers Erik Brynjolfsson, Bharat Chandar and Ruya Chen captured this dynamic. AI automates book learning but not the judgment that comes from experience. It substitutes for new graduates. It complements veterans.
Wages tell another story. They rose across AI-exposed sectors. Computer systems design saw average weekly wages jump 16.7 percent since late 2022, outpacing the national 7.5 percent gain. Top AI-exposed industries posted 8.5 percent growth. No clear negative link appeared between an occupation’s AI exposure and its wage gains. In fact, the experience premium, the pay gap between veterans and rookies, correlates positively with AI exposure. Median premium sits at 40 percent. Higher exposure often means bigger rewards for those who stay. But that leaves newcomers on the outside looking in.
Broader forecasts paint a complex picture. Boston Consulting Group estimates that 50 to 55 percent of U.S. jobs will face reshaping by AI within two to three years. Only 10 to 15 percent appear vulnerable to outright elimination after accounting for demand growth and task complementarity. Many roles will shift. Workers will handle new expectations. New tasks will emerge. Yet the transition carries risks. Brookings Institution researchers examined adaptive capacity among the 37.1 million workers in the top quartile of AI exposure. About 26.5 million show above-average ability to handle job changes thanks to skills, savings or local opportunities. Still, 6.1 million, mostly in clerical and administrative work and 86 percent women, lack that buffer. They face high exposure paired with low adaptability. (Brookings Institution)
Anthropic’s recent framework tested early labor data against AI exposure measures. It found no broad rise in unemployment for highly exposed workers since late 2022. Hiring for younger workers in those fields, however, slowed. BLS projections suggest occupations with higher observed exposure will grow more slowly through 2034. The signals point to a quiet pullback at the entry point rather than mass layoffs.
Tech companies provide visible examples. Layoffs tied to AI adoption mounted. Some 78,000 tech jobs vanished in the first half of 2025 alone, many linked to efficiency drives. Early 2026 brought another 32,000 cuts in the sector. Firms pour billions into AI infrastructure. They reduce headcount in functions now partly automated. Productivity climbs. Demand for certain outputs expands. But the headcount math changes.
So what does this mean for industries? Office support, customer service and administrative functions sit in the crosshairs. These tasks involve routine processing and standardized decisions. AI handles them at scale. Marketing, banking, travel and retail show moderate pressure too. Blue-collar roles, by contrast, added about one million more positions than white-collar work over recent years. Manual labor grew modestly as office employment edged down. The shift reverses decades of white-collar job leadership.
Optimists point to new opportunities. AI creates demand for skills in model training, data curation, system oversight and prompt engineering. Infrastructure spending on data centers, chips and power generation generates construction and engineering roles. Yet those positions demand different qualifications than the ones displaced. Retraining takes time. Geographic mismatches appear. University towns and midsized markets in the Mountain West and Midwest hold concentrations of vulnerable clerical workers with few local alternatives.
IMF analysis adds nuance. Demand for new AI-related skills links to lower employment in occupations with high exposure and low complementarity to the technology. In regions with rising AI skill postings, employment in those vulnerable roles ran 3.6 percent lower after five years. Non-AI new skills, however, associated with higher overall employment. The distinction matters. AI skills themselves concentrate among ICT and STEM backgrounds. AI-user skills spread more widely. Yet the net employment effect so far remains muted or negative for some groups, especially youth and middle-skill white-collar workers.
Policy makers and business leaders watch closely. Short-term displacement could linger if new job creation lags. Goldman Sachs researchers earlier projected that AI might expose 300 million jobs globally to automation. In the U.S., a front-loaded transition might lift unemployment noticeably before productivity gains spread. Recessions would amplify the effects. But augmentation also fuels expansion. Lower costs can open new markets. Human judgment, creativity and accountability retain value. Roles requiring physical presence or complex interpersonal decisions often gain.
The data remains early. Monthly net losses of 16,000 jobs matter but haven’t yet produced economy-wide unemployment spikes. Total employment still grows. Wages in exposed fields hold or rise. The real test lies ahead as adoption scales. Companies that once experimented with AI now integrate it into core operations. Efficiency metrics improve. Headcount reviews follow. Young workers bear the initial brunt because their tasks map most cleanly to what large language models do best.
Experience, it turns out, buys protection. Tacit knowledge resists easy codification. Veterans who combine domain expertise with AI tools become more valuable. New entrants must prove they bring something machines cannot yet match. That raises the bar. It pulls up the ladder for some. And it forces a reckoning with how societies prepare workers for an economy where routine cognitive work loses ground.
Recent analyses reinforce the mixed signals. A BCG study from April 2026 highlights that reshaping will outpace elimination for most jobs. (BCG) Morgan Stanley and Goldman Sachs both concluded AI added only about 0.1 percentage point to the unemployment rate at most, with substitution effects partly offset by augmentation. Yet the entry-level pain stays real. Surveys from FlexJobs in April 2026 found 75 percent of workers reported no personal job changes from AI so far. Concern runs higher at 42 percent. The gap between perception and current statistics may close as more firms move beyond pilots.
One thing feels clear. The era of waiting for AI’s labor market effects has ended. Early numbers confirm displacement in targeted areas. They also show adaptation, wage resilience and new demand in complementary fields. The coming years will test whether productivity gains translate into broad opportunity or concentrated gains for those positioned to ride the wave. Businesses must weigh task redesign, reskilling and hiring strategies against the visible efficiency numbers. Workers, especially those starting out, face a market that values proven judgment over fresh credentials alone. The data has spoken. Now the adjustments begin.
WebProNews is an iEntry Publication