AI and Jobs Across 50 Countries: Why Technology Alone Cannot Deliver Shared Progress

The findings describe patterns across those observations without establishing that AI caused changes in employment or development.

Buying better technology is easier than making sure people benefit from it, especially when workers need new skills and employers are changing how jobs get done. A study covering 50 countries explores this gap between being ready for artificial intelligence and turning that readiness into better lives. Its clearest findings connect stronger national AI capacity with lower unemployment and a larger share of people in work.

Gustavo Pessoa and colleagues examine these relationships in "Artificial intelligence diffusion, adaptive Human-Capital Transformation, and unequal development: evidence from a 50-country panel," published in Frontiers in Artificial Intelligence. Their research draws on 229 country-year observations from 2020–2024, covering economies with different incomes, education systems and technological resources. The findings describe patterns across those observations without establishing that AI caused changes in employment or development.

The paper places people's ability to adapt at the centre of the discussion, looking beyond qualifications or access to digital tools. A worker may need opportunities to refresh existing knowledge, combine skills in unfamiliar ways or move into a different occupation. Countries face a similar challenge in connecting technological investment with learning, employment support and institutions that help people navigate change.

What the researchers measured

The team built five measures covering AI capacity, educational and digital foundations, government readiness, labour-market exposure and broader development outcomes. AI capacity combined technology-sector strength with data and infrastructure, measuring the conditions that support AI deployment rather than actual use inside workplaces. The human-capital measure combined higher-education enrolment with internet use, which cannot directly reveal how well someone understands AI or adapts to a changing job.

Government readiness represented enabling conditions, and the share of employment in services served as a rough indicator of potential exposure to changing tasks. The development measure combined economic growth, labour productivity, labour-force participation and life expectancy, giving the researchers a broader view than economic output alone. Their analysis accounted for lasting differences between countries and shared annual shocks, with additional checks using alternative index weights and repeated sampling.

The authors describe "adaptive human capital" as the ability to preserve, renew, recombine and redeploy human capabilities. Health helps sustain that ability, lifelong learning supports renewal, and opportunities to apply skills help people move through technological change. Their proposed Adaptive Human Capital Index remains a framework for future measurement because comparable annual data were insufficient to calculate it.

Employment showed the clearest connection

Within countries, increases in AI capacity were associated with lower overall unemployment, lower youth unemployment and a higher employment-to-population ratio. These relationships held after accounting for income and other conditions, including checks designed to reduce false discoveries when testing several relationships. The findings do not establish that AI creates jobs automatically or reveal which occupations gained work and which lost it.

AI capacity was positively associated with the study's educational and digital measure, offering evidence that these foundations developed together during the period examined. Connections with life expectancy, labour-force participation and the combined development measure were less convincing under stricter testing, and the broader development relationship weakened after income was included. The analysis found no reliable association with economic growth or labour productivity, and it did not establish that stronger educational and digital foundations amplified AI's development benefits.

Similar AI capacity, very different realities

The researchers divided the 50 countries into four groups using AI capacity and unemployment, with the sample's median unemployment rate of 4.545% as the dividing line. Fourteen countries, including Japan, Germany and the United States, combined above-average AI capacity with lower unemployment; another 12, including Canada, China and France, combined above-average capacity with higher unemployment. Average youth unemployment reached about 17.5% in the second group, compared with 8.7% in the first.

Eleven countries, including India, Indonesia and Mexico, paired below-average AI capacity with lower unemployment. Another 13, including Brazil and South Africa, faced both below-average AI capacity and higher unemployment, recording the weakest average development score among the four groups. These categories describe positions within this particular sample, so low unemployment alone should not be read as evidence of successful AI adoption or good-quality jobs.

A separate comparison of AI capacity with educational and digital foundations placed 24 countries above average on both measures and 15 below average on both. Two countries had stronger technology than human-capital foundations, and nine showed the reverse pattern, suggesting a risk that existing capabilities were not being fully supported by technological capacity. Comparisons between countries with similar AI scores found no single measured factor that consistently explained their different development outcomes.

Turning technological potential into everyday opportunity

The paper explores possible paths for 2026–2027 using lower, middle and upper rates of change drawn from the observed AI and human-capital data. The lower joint path produced declining development scores, and the middle and upper paths produced improvements, with some countries moving between groups. These are conditional illustrations based on assumptions, not forecasts or evidence of a proven reinforcing effect between technology and skills.

The authors recommend preserving job quality and continued learning where AI capacity and employment are relatively strong, and improving retraining, job matching and transition support where unemployment remains higher. Countries with lower AI capacity and comparatively low unemployment need technological and institutional investment that supports existing employment, with broader investment in infrastructure, skills, employment services and social protection where both dimensions are weak. These priorities connect AI investment with education, decent work and reduced inequality, giving governments reasons to look beyond technology rankings.

The study covers a short period and a deliberately selected sample, limiting how far its findings can represent the world or describe lasting changes. National averages cannot show how gains and losses are distributed across individual workers, and the measures cannot directly capture AI competence, workplace displacement or successful retraining. Judging progress requires looking at people's opportunities to learn, use their capabilities and find decent work alongside a country's technological capacity.