Executive Summary
The arrival of ChatGPT in late 2022 triggered a wave of predictions about the future of work that ranged from catastrophic to utopian. Goldman Sachs warned that 300 million jobs globally could be affected by generative AI. The World Economic Forum projected the creation of 97 million new roles. McKinsey estimated that up to 30 percent of hours worked in the United States could be automated by 2030.
Now, more than three years into the generative AI era, we have enough real-world data to begin separating signal from noise. The picture that emerges is neither the mass unemployment crisis that some feared nor the productivity paradise that others envisioned. It is something more complex, more uneven, and in many ways more interesting than either extreme.
Our analysis draws on Bureau of Labor Statistics employment data, Challenger, Gray and Christmas layoff tracking reports, LinkedIn Economic Graph hiring data, Federal Reserve surveys of business conditions, and proprietary data from our AI Job Impact Scanner. The findings suggest that the AI revolution is real but proceeding along a different trajectory than most analysts anticipated.
The Displacement Picture
The headline finding is that AI-related job displacement has been significant but far more targeted than the sweeping predictions of 2023 suggested. According to BLS data through February 2026, total nonfarm employment in the United States stands at approximately 159.2 million, reflecting net growth of 4.1 million jobs since January 2023. The unemployment rate sits at 3.9 percent, slightly above its 2023 low of 3.4 percent but well within historical norms.
These aggregate numbers, however, obscure important shifts happening beneath the surface. Challenger, Gray and Christmas data shows that employer-cited "technological redundancy" as a reason for layoffs has increased 340 percent since 2022. In 2025 alone, firms announced approximately 185,000 layoffs explicitly attributed to AI and automation, up from roughly 42,000 in 2023. The industries most affected include financial services, media and publishing, customer service operations, and back-office administrative functions.
The pattern of displacement has been notably different from what many forecasters expected. Rather than eliminating entire occupations wholesale, AI has primarily reduced headcount within specific functional areas. A corporate legal department that employed twelve paralegals in 2023 might now employ seven, with the remaining staff handling a larger volume of work with AI assistance. A newsroom that maintained twenty copy editors might now operate with eight, supplemented by AI-powered editing tools.
This task-level displacement, as opposed to occupation-level elimination, is a critical distinction. BLS occupational employment data shows that very few job categories have experienced outright decline attributable primarily to AI. Instead, growth rates have decelerated in affected categories, and new hiring has slowed even as existing workers remain employed. Entry-level positions have been disproportionately affected, as organizations have found that AI tools can handle many of the tasks traditionally assigned to junior staff.
The Challenger data reveals a geographic concentration as well. Metropolitan areas with heavy concentrations of information work, particularly San Francisco, New York, Chicago, and Austin, have seen the most AI-related displacement. Regions with economies centered on healthcare, construction, and skilled trades have been largely unaffected.
The Creation Side
If displacement has been more targeted than predicted, job creation has been more robust and more diverse than many anticipated. LinkedIn Economic Graph data indicates that job postings requiring AI-related skills have grown by approximately 280 percent since January 2023. More significantly, an entirely new category of roles has emerged that did not exist in any meaningful form before the generative AI wave.
The most visible new category is AI operations and integration. Organizations that have adopted AI tools have discovered that deploying them effectively requires dedicated human oversight, configuration, and management. This has generated substantial demand for roles such as AI integration specialists, prompt engineers, AI output quality analysts, and machine learning operations engineers. LinkedIn data shows approximately 74,000 active job postings in the United States for roles that did not have standardized job titles before 2024.
A second category of job creation has come from the expansion of industries that AI has made more productive or more accessible. AI-powered design tools have not eliminated graphic designers; they have enabled a much larger number of small businesses and entrepreneurs to invest in professional-quality visual content, expanding the overall market for design services even as the per-project labor requirement has declined. Similar dynamics are playing out in software development, content creation, and data analysis.
The industry-level data we track shows this pattern most clearly in healthcare, where AI diagnostic and administrative tools have not reduced clinical hiring but have enabled expansion of services into underserved areas. Telehealth platforms using AI triage have created demand for remote clinical coordinators, AI-assisted care navigators, and health data integrity specialists, none of which existed at meaningful scale before 2024.
Federal Reserve Beige Book reports from regional banks consistently note that businesses adopting AI are, on average, adding headcount in new functional areas even as they reduce it in traditional ones. The net employment effect varies by firm size: large enterprises with more than 5,000 employees have tended toward modest net reduction, while small and mid-size firms have tended toward net growth as AI tools have enabled them to compete in markets previously dominated by larger players.
Wage Effects
The wage picture is where the AI economy reveals its most significant inequalities, and where the data demands the most careful interpretation. BLS earnings data through the fourth quarter of 2025 shows that median real wages across all occupations have grown at approximately 1.8 percent annually, roughly in line with historical norms. But this average conceals a divergence that is accelerating.
Workers with demonstrable AI skills command a significant premium. Analysis of compensation data from multiple sources indicates that professionals who can effectively work with AI tools, meaning not just use them but integrate them into complex workflows and evaluate their outputs critically, earn 18 to 32 percent more than peers in equivalent roles without those skills. This premium is highest in software engineering, financial analysis, marketing strategy, and legal services.
Conversely, workers in roles where AI has reduced the skill barrier have seen wage stagnation or decline. Junior copywriters, entry-level data analysts, basic bookkeepers, and first-tier customer service representatives have experienced real wage declines of 5 to 12 percent as the supply of workers capable of performing these roles, augmented by AI, has expanded faster than demand.
The Brookings Institution has documented a widening gap between what they term "AI-complementary" workers and "AI-substitutable" workers. The former group, whose skills are enhanced by AI tools, has seen the fastest wage growth in over two decades. The latter group, whose core tasks can be partially or fully performed by AI systems, faces increasing downward pressure on compensation. This bifurcation follows historical patterns observed during previous technological transitions, but the speed of divergence is unprecedented.
Geographic wage effects mirror the displacement patterns. In technology hubs, AI-skilled workers are commanding premium compensation while displaced workers face intense competition. In regions less affected by AI adoption, wage structures have remained more stable but growth has been slower overall.
Industry Variations
The impact of AI on employment is not a single story but a collection of very different stories playing out across sectors. Our industry tracking reveals several distinct patterns.
Technology and Software. The sector that created AI has been among those most affected by it. Software development employment has grown modestly, approximately 6 percent since 2023, but the composition has shifted dramatically. Demand for senior engineers with AI expertise has surged while entry-level coding positions have contracted. Companies report that AI-assisted development tools have increased per-developer productivity by 40 to 55 percent, allowing teams to accomplish more with fewer junior hires. GitHub data shows that AI-generated code now accounts for roughly 35 percent of all code committed to enterprise repositories.
Financial Services. Banks, insurance companies, and asset managers have been among the most aggressive adopters of AI, and the employment effects are substantial. Back-office processing, compliance monitoring, and routine financial analysis have seen significant headcount reductions. JPMorgan Chase, Bank of America, and other major institutions have publicly discussed reducing analyst classes by 20 to 30 percent. However, new roles in AI risk management, algorithmic compliance, and AI-augmented advisory services have partially offset these reductions.
Healthcare. This sector has seen the most positive net employment effect from AI adoption. Administrative efficiency gains have been real but have largely been absorbed through attrition rather than layoffs, as the sector continues to face chronic staffing shortages. AI diagnostic tools, rather than replacing clinicians, have expanded the range of services that existing staff can provide. The sector has added approximately 890,000 jobs since 2023, with AI-related roles representing a growing share of new positions.
Media and Creative Industries. Perhaps no sector has experienced more disruption relative to its size. Generative AI has fundamentally altered the economics of content production, and employment in traditional media roles, including journalism, copywriting, graphic design, and video production, has declined measurably. Challenger data shows media and entertainment layoffs citing AI increased by over 400 percent between 2023 and 2025. Yet the creative economy has also expanded into new territory, with demand for AI-assisted creative direction, synthetic media production, and human-AI creative collaboration growing rapidly.
Manufacturing and Skilled Trades. These sectors remain among the least affected by generative AI specifically, though robotics and industrial automation continue their multi-decade trajectory. The skilled trades, including electricians, plumbers, HVAC technicians, and construction workers, have experienced virtually no AI-related displacement and continue to face labor shortages. AI planning and logistics tools have improved productivity without reducing headcount.
What Experts Got Right and Wrong
The track record of AI employment predictions made in 2023 and early 2024 is decidedly mixed, and examining where the experts went astray offers useful lessons for calibrating future expectations.
What they got right. The broad consensus that AI would affect knowledge work more than physical work has been confirmed by the data. White-collar information processing roles have seen the most significant changes, while hands-on trades and healthcare delivery have been largely insulated. The prediction that AI would primarily augment rather than replace workers has also proven generally accurate at the occupation level, even as task-level displacement has been real and consequential. McKinsey's framework of analyzing automation at the task level rather than the job level has proven to be the most analytically useful approach.
What they got wrong. The most common error was overestimating the speed of adoption. Enterprise AI deployment has proceeded far more slowly than the breathless announcements of 2023 suggested. Surveys by Deloitte and Accenture consistently show that only 25 to 35 percent of large enterprises have moved beyond pilot programs into scaled AI deployment. Integration challenges, data quality issues, regulatory uncertainty, and organizational resistance have all slowed the pace of adoption.
A second major miss was underestimating the new job creation effect. Most early analyses focused heavily on which jobs would be eliminated without adequately modeling the new categories of work that AI would generate. The historical pattern of technology creating more jobs than it destroys, while not guaranteed to hold indefinitely, has operated broadly as expected through the first three years of the generative AI era.
A third error was treating AI capabilities as static. The models available in early 2023 had significant limitations in reliability, accuracy, and contextual understanding. Many displacement predictions were implicitly based on extrapolations from those early capabilities that have not fully materialized. While AI systems have improved substantially, they remain less reliable and more narrowly capable than many forecasters assumed.
Looking Ahead
The data we have examined supports several forward-looking observations, though we offer these with appropriate humility given the forecasting track record we have just critiqued.
First, AI's impact on employment will likely accelerate over the next two to three years as enterprise adoption moves from pilot to scale. The 65 to 75 percent of large organizations still in early stages represent a substantial wave of deployment that will generate both displacement and creation effects.
Second, the wage bifurcation between AI-complementary and AI-substitutable workers is likely to widen before it stabilizes. Workers who invest in AI literacy and learn to work effectively with these tools will continue to command premium compensation. Those who do not will face increasing competitive pressure. Our AI Job Impact Scanner provides individualized assessments of where specific roles fall on this spectrum.
Third, policy responses will increasingly shape outcomes. To date, government intervention has been limited primarily to AI safety regulation and modest investments in retraining programs. As displacement effects become more visible and more concentrated in specific communities and demographics, pressure for more substantive policy responses, including adjustment assistance, education reform, and potentially new models of work, will intensify.
Fourth, the industries that have been least affected so far, including healthcare, skilled trades, and education, are not immune to future disruption. Advances in robotics, embodied AI, and multimodal systems could extend automation into physical and interpersonal domains that have so far remained largely human.
The state of AI and jobs in 2026 is neither crisis nor utopia. It is a labor market in transition, with real winners and real losers, significant uncertainty, and a pace of change that demands continuous monitoring and evidence-based analysis. That is precisely what this publication exists to provide.