Will AI create or take away more job opportunities?

AI will both create and take away job opportunities, with the net long-term effect likely positive or roughly neutral in terms of total employment (following historical patterns of technological change), but with substantial short- to medium-term disruption, occupational shifts, and uneven impacts. The outcome depends heavily on how AI is deployed (pure substitution vs. augmentation of humans), the pace of adoption, complementary investments in skills/education, economic growth, and policy responses. Mass technological unemployment is widely viewed as unlikely by most economists and institutions, though transition challenges could be severe for many workers.[1]

Current Evidence (as of mid-to-late 2020s data)

Empirical studies so far show AI simultaneously destroying and creating jobs, often with a modest net drag or neutral effect on overall employment, rather than widespread layoffs:

•Goldman Sachs Research finds a modest net drag on US labor markets: AI reduced monthly payroll growth by roughly 16,000 jobs over the past year (raising unemployment by about 0.1 percentage point). Substitution effects reduce jobs in high-risk occupations, but augmentation (where AI boosts human productivity) increased monthly payroll growth by about 9,000 jobs. Negative effects hit younger/less-experienced workers harder. Productivity gains can lower costs and boost demand enough to raise net employment in augmented roles (an example of Jevons paradox).[2]

•Analyses separating substitution from complementarity (e.g., CEPR and related work) show substitutable tasks/occupations growing ~3.3 percentage points slower per year since around 2021, while complemented ones grow ~3.3 points faster. In headcount terms (US-focused estimates), this can mean roughly 1.5 million jobs/year “removed” via slower growth vs. 0.7 million added—net depends on economic composition. Adjustment often occurs via reduced hiring rather than mass firing, disproportionately affecting the young.[3]

•RAND and Stanford SIEPR research indicates AI has so far been more associated with employment increases than decreases at adopting firms. Overall unemployment impacts remain small; rates in highly AI-exposed occupations have not risen faster than in less-exposed ones. Many firms report little or no headcount change from AI, and some high-spending adopters grow employment. Actual workforce reductions have fallen short of prior expectations (e.g., McKinsey surveys: ~14% of firms reported AI-related declines vs. higher shares that had predicted them).[4]

•PwC’s Global AI Jobs Barometer links greater AI exposure to headcount growth, not decline, as firms redesign operations. It describes a “two-track” market: certain professionalized roles grow faster with higher wage growth.[5]

Entry-level and certain white-collar/routine cognitive roles (e.g., some admin, coding, customer service, clerical) show softer hiring or declines in postings in exposed fields, while demand rises for AI-related specialists, oversight, and complementary skills.[6]

Projections for the Coming Years

Major reports generally forecast net job creation or limited net loss, alongside high churn:

•World Economic Forum Future of Jobs Report 2025: Macro trends (including technological change as the biggest driver of both creation and displacement, plus green transition, demographics, etc.) are expected to drive a net creation of 78 million jobs globally by 2030. Overall: ~14% job growth offset by ~8% decline. Tech drives many of the fastest-growing and declining roles. Over two-thirds of employers plan to hire for AI-specific roles, even as ~40% foresee workforce adjustments. Skills will shift significantly (~40% of core skills changing).[7]

•McKinsey: In the US by 2035, AI/automation could cut demand for ~36 million jobs while creating ~41 million elsewhere (net positive). However, roughly 11 million workers (~7% of the workforce; range 6–16 million) may need to switch occupations entirely—far above historical rates of occupational change. Lower-wage workers face higher risks of needing big shifts; growth areas include healthcare, construction, and management. The core challenge is mobility and retraining, not absolute scarcity of jobs. Only a minority of displaced workers have easy paths into growing roles without substantial reskilling. Earlier McKinsey work and history emphasize that tech creates more jobs long-term than it destroys.[8]

•Other views (IMF, various economists): Significant shares of jobs (e.g., 40% globally, higher in advanced economies) will be affected (augmented or partially automated), but full displacement is lower. Low-wage roles can face higher substitution risk for core tasks in some models, potentially worsening inequality, while high-skill/complex roles often see more augmentation.[9]

Productivity boosts from AI are expected to generate new demand, new industries/roles (many currently unforeseeable, as with past tech), and expansion in areas like healthcare, education, green tech, and creative/oversight work.

Historical Context

Technological revolutions (industrial, electricity, computers, internet) have repeatedly caused large-scale displacement in specific sectors (e.g., agriculture from ~60% of US employment in 1850 to <5%; manufacturing declines; handloom weavers, telephone operators) while creating far more jobs overall and raising living standards. About 60% of current US jobs are in categories that didn’t exist in 1940.[10]

However, net effects are not automatic or immediate:

•Short-term pain can be prolonged for affected groups (skill mismatches, geographic issues, age).

•Some research (e.g., Autor et al. at MIT) finds that since ~1980, automation has replaced more US jobs than augmentation created, contributing to polarization and slower gains for middle/lower-skill workers—though total employment still grew with the economy, and new task creation continued.[10]

•New jobs often emerge outside the tech sector itself and require adaptation. “This time is different” arguments (AI’s cognitive capabilities) have merit for speed and breadth of impact on white-collar work, but macroeconomic job creation via productivity/wealth effects has held historically.[1]

Key Uncertainties and What Matters Most

•Augmentation vs. replacement: AI that makes workers more productive (and expands output/demand) creates more jobs than pure automation of fixed tasks.

•Pace and diffusion: Adoption is uneven; full effects take years/decades. Infrastructure (data centers, energy) and complementary innovations matter.

•Who is affected: Younger/entry-level workers, certain routine cognitive/admin roles, and lower-wage workers in some analyses face higher near-term risks. High-skill, interpersonal, creative, physical, or judgment-heavy roles often benefit more from complementarity. Inequality could rise without intervention.

•New work: Roles in AI development/governance, human-AI collaboration, prompt/engineering oversight, ethics, maintenance of complex systems, and entirely new domains enabled by cheaper intelligence/computation.

•Policy and societal choices: Education/reskilling (lifelong learning), safety nets, incentives for job-creating AI uses (e.g., in care sectors), competition policy, and managing transitions determine whether benefits are widely shared. Skills gaps are already a top barrier cited by employers.

In summary, AI is more likely to reshape and reallocate job opportunities than to cause a permanent net reduction in total employment. It will take away many existing roles or tasks (especially routine ones) while creating others—often better-paying or in expanding sectors—and boosting productivity that supports overall demand for labor. The net balance has historically favored creation over long horizons, and current projections (WEF, McKinsey) lean that way, but the transition will involve significant churn, requiring proactive adaptation by individuals, firms, and governments. The biggest risks are not “no jobs,” but mismatched skills, unequal access to opportunities, and inadequate support during shifts. Continuous monitoring of labor data, investment in human capital, and flexible policies will be essential.

Citations (10)

1weforum.orgNonprofits and think tanks
https://www.weforum.org/stories/technological-innovation/why-ai-will-not-lead-to-a-world-without-work
Backs: “Mass technological unemployment is widely viewed as unlikely by most economists and institutions, though transition challenges could be severe for many workers.”
2goldmansachs.comCompanies
https://www.goldmansachs.com/insights/articles/the-jobs-ai-is-likely-to-boost-and-those-it-may-disrupt
Backs: “Productivity gains can lower costs and boost demand enough to raise net employment in augmented roles (an example of Jevons paradox).”
3cepr.orgNonprofits and think tanks
https://cepr.org/voxeu/columns/hiding-mean-ai-adding-jobs-and-removing-them-same-time
Backs: “Adjustment often occurs via reduced hiring rather than mass firing, disproportionately affecting the young.”
4rand.orgNonprofits and think tanks
https://www.rand.org/pubs/commentary/2025/10/ai-is-making-jobs-not-taking-them.html
Backs: “Actual workforce reductions have fallen short of prior expectations (e.g., McKinsey surveys: ~14% of firms reported AI-related declines vs. higher shares that had predicted them).”
5pwc.comCompanies
https://www.pwc.com/gx/en/issues/artificial-intelligence/job-barometer/2026/2026-global-ai-jobs-barometer-full-report.pdf
Backs: “It describes a “two-track” market: certain professionalized roles grow faster with higher wage growth.”
6hbs.eduAcademic
https://www.library.hbs.edu/working-knowledge/ai-hasnt-spurred-widespread-layoffs-yet-why-not
Backs: “Entry-level and certain white-collar/routine cognitive roles (e.g., some admin, coding, customer service, clerical) show softer hiring or declines in postings in exposed fields, while demand rises for AI-related specialists, oversight, and complementary skills.”
7weforum.orgNonprofits and think tanks
https://www.weforum.org/stories/jobs-and-the-future-of-work/future-of-jobs-report-2025-whats-shaping-the-future-of-the-global-workforce
Backs: “Skills will shift significantly (~40% of core skills changing).”
8yahoo.comNews and mediaquality 0.69
https://finance.yahoo.com/technology/ai/articles/ai-create-more-jobs-kills-080000226.html
Backs: “Earlier McKinsey work and history emphasize that tech creates more jobs long-term than it destroys.”
9bis.orgGovernment
https://www.bis.org/publications/working-paper-1207-rise-generative-ai-modelling-exposure-substitution-and-inequality-effects-us-labour-market
Backs: “Low-wage roles can face higher substitution risk for core tasks in some models, potentially worsening inequality, while high-skill/complex roles often see more augmentation.”
10mit.eduAcademic
https://news.mit.edu/2024/does-technology-help-or-hurt-employment-0401
Backs: “About 60% of current US jobs are in categories that didn’t exist in 1940.”