AI Impact on Jobs | ILO G20 Employment Analysis

Understanding AI Impact on Jobs in the G20 Economy

The rapid proliferation of artificial intelligence, particularly Generative AI, is fundamentally reshaping how we think about work across the world’s largest economies. A landmark paper from the International Labour Organization (ILO), prepared for the G20 Framework Working Group, provides the most comprehensive assessment to date of how AI adoption is transforming employment patterns, skill requirements, and job quality across advanced and developing nations alike.

The ILO’s analysis delivers a nuanced and data-driven perspective that challenges both utopian and dystopian narratives about AI’s impact on jobs. Rather than triggering mass unemployment, the evidence suggests AI is more likely to augment human capabilities and enhance productivity in many roles. However, this encouraging headline masks substantial disparities—certain occupations, demographic groups, and regions face significantly higher exposure to AI-driven transformation, demanding urgent and targeted policy responses.

For policymakers, business leaders, and workers navigating this technological transition, understanding these patterns is essential. The paper builds on the ILO’s pioneering 2023 global estimates of occupational exposure to GenAI—which found over 30% of global employment potentially exposed—by introducing a refined methodology with four progressive gradients of impact. This updated framework enables more precise identification of who is most at risk, who stands to benefit, and what interventions can bridge the gap. Explore more analyses of how AI and technology are transforming industries in our interactive library.

Four Gradients of AI Exposure Across Occupations

The ILO’s refined methodology classifies occupational exposure to GenAI into four distinct gradients, moving beyond the binary automation-versus-augmentation framework to capture the full spectrum of AI’s potential impact on different types of work.

Gradient 1 encompasses occupations with low overall exposure but high variability across constituent tasks. While certain tasks within these roles may show substantial automation potential, the majority continue to require distinctly human capabilities—think healthcare workers or educators whose roles blend routine documentation with complex interpersonal judgment.

Gradient 2 reflects moderate exposure with heterogeneous task composition. Some tasks are highly exposed while others remain largely unaffected, creating uneven effects within the same occupation. A marketing manager, for instance, might see AI transform data analysis tasks while leaving strategic client relationship work untouched.

Gradient 3 captures occupations where a significant share of tasks is consistently exposed to GenAI functionalities, suggesting escalating automation risks and a pressing need for worker adaptation. Financial analysts and legal researchers fall into this category, where AI tools can perform substantial portions of their daily work.

Gradient 4 represents the highest and most uniform levels of task exposure, where the preponderance of tasks exhibits strong alignment with GenAI capabilities. Clerical workers—data entry operators, administrative assistants, and bookkeepers—dominate this gradient, with 24% of clerical tasks categorized as highly exposed and another 58% as medium-level exposed.

Approximately one quarter of total global employment falls within one of these four exposure gradients. In the G20 specifically, 25% of total employment is classified within exposure gradients, closely mirroring the global pattern. This analytical precision allows organizations and governments to develop targeted interventions rather than applying one-size-fits-all solutions.

Gender Disparities in AI Automation Risk

Perhaps the most striking finding of the ILO analysis is the pronounced gender gap in AI exposure. Globally, 28% of female employment is situated within exposed occupations compared to 21% for male employment—a 7 percentage point gap that translates to millions of workers facing differential levels of technological disruption.

The disparities intensify at higher exposure levels. Women are more concentrated in the top two most exposed gradients (Gradients 3 and 4), with 10% of global female employment in these categories compared to just 6% for men. In G20 economies, the female share in the two most exposed gradients reaches 13%, more than double the male share of 6%.

Geographic context amplifies these gender effects. In advanced economies, the gender gap in GenAI-exposed employment reaches 13 percentage points—compared to 6 percentage points in emerging markets and developing economies. The European Union shows an even wider gap of 15 percentage points, while the African Union’s gap is 5 percentage points. These patterns reflect persistent occupational sex segregation, particularly the concentration of women in clerical, financial, and customer service roles that are both disproportionately represented in higher-income economies and among those most susceptible to GenAI transformation.

This gendered dimension demands that AI workforce transition policies explicitly address the needs of women workers. Retraining programs, social protection systems, and active labor market policies must account for the reality that women face higher automation risk not because of inherent skill deficits, but because of historical patterns of occupational segregation that now intersect with the specific capabilities of generative AI technology.

AI Impact on Young Workers and Entry-Level Jobs

Beyond gender, the ILO’s analysis reveals a generational dimension to AI disruption that carries profound implications for workforce development and economic mobility. Young workers, who typically fill entry-level positions, face a distinctly different—and potentially more threatening—exposure profile than older, more experienced colleagues.

ILO research finds that young people are more likely to work in occupations at greater risk of automation. In the United Kingdom, research shows that entry-level and part-time positions could be the most exposed in the early adoption stages of AI. A parallel study in the United States suggests that while more senior roles may experience a productivity boost from AI augmentation, demand for entry-level jobs may erode.

This dynamic threatens what the ILO describes as the “traditional career ladder and skill development pathway.” Entry-level roles have historically served dual purposes—providing income while building the experiential foundation for career advancement. If AI displaces these positions, it creates a fundamental gap in how the next generation acquires workplace skills and professional networks.

The implications are particularly acute for Africa, where the ILO projects that at least 350 million young people will be of working age over the next two decades. By 2050, one in three young people globally aged 1–24 will be of sub-Saharan African origin. This demographic surge represents either an immense opportunity for a demographic dividend that could fuel economic growth and innovation, or a significant development challenge if sufficient productive and decent employment opportunities are not created in an increasingly AI-transformed labor market.

Digital Employment Intensity Across Income Levels

To contextualize AI exposure, the ILO examines the current digital intensity of employment across different economic environments—revealing stark disparities that shape how AI will realistically impact different regions. Using a taxonomy that classifies sectors into four categories of digital intensity (high, medium-high, medium-low, and low), the analysis shows fundamentally different starting points for digital transformation.

In advanced economies and the EU, employment is relatively evenly distributed: approximately 50% of employment is in sectors classified as high or medium-high in digital intensity, with the remaining half in medium-low and low-intensity sectors. This balance means AI disruption will be felt broadly but with existing infrastructure to manage the transition.

In contrast, G20 emerging markets and developing economies show a markedly skewed distribution, with less than 30% of workers in digitally intensive sectors and the majority concentrated in low and medium-low digital-intensive sectors. The African Union exhibits even lower digital employment shares. These structural differences mean that while developing economies face lower immediate AI exposure, they also have fewer pathways to capture AI’s productivity-enhancing benefits. Discover how different sectors are responding to digital transformation through our curated library of industry analyses.

Women’s employment in digitally intensive sectors is significantly lower than men’s across all income levels, with particularly low shares in emerging markets and the African Union. This compounds the gender vulnerability identified in AI exposure analysis, creating a double disadvantage: women are overrepresented in the occupations most vulnerable to AI automation while simultaneously underrepresented in the digitally advanced sectors best positioned to benefit from AI augmentation.

From AI Exposure to Real Employment Effects

Moving from theoretical exposure to empirical evidence, the ILO presents findings on how exposure to 40 emerging digital technologies—including AI—actually affects employment across 50 countries at different income levels. Using an instrumental variable shift-share approach, the analysis provides causal evidence rather than mere correlations.

The results reveal a surprising and important pattern: the largest employment-to-population gains from exposure to emerging digital technologies occurred in low-income countries, with an average increase of 1.05 percentage points. Upper-middle-income and high-income countries followed with more modest gains of 0.13 and 0.12 percentage points respectively, while lower-middle-income economies experienced negligible change.

At the subregional level, the Arab States showed the most substantial gains (+1.75 percentage points), followed by North America (+0.92) and North Africa (+0.54). In contrast, some parts of Asia experienced negative employment effects, particularly East Asia (−0.23) and South-East Asia (−0.15). These patterns underscore what the ILO calls the “heterogeneity in digitalization’s labour market effects”—supporting the notion that lower-income and less-automated regions benefit more consistently from technology adoption, while more digitally mature economies face varied or even adverse outcomes where job displacement from automation is more pronounced.

The OECD’s complementary research on AI and labor markets corroborates these findings, noting that the net employment effects of digital transformation depend heavily on local economic structures, institutional frameworks, and the speed of skills adaptation.

The Role of Skills in Navigating AI Transformation

Skills emerge as the single most important mediating factor in determining whether AI exposure translates into job loss or job enhancement. The ILO’s analysis of how occupational skill composition interacts with digital technology exposure reveals that the right skill mix can transform technological disruption into economic opportunity—but the effectiveness of different skills varies dramatically across income levels.

Across all country income groups, cognitive and managerial skills consistently amplify the positive effects of digital exposure on employment. Core and sophisticated cognitive abilities, computer literacy, writing, financial skills, and project and process management all serve as employment buffers and growth catalysts. These effects are strongest in low- and lower-middle-income countries, where such skills are strongly associated with employment resilience and job creation.

The role of advanced technical skills—software development, machine learning, and AI—tells a more nuanced story. In high-income countries, these skills are negatively associated with employment outcomes, likely reflecting job polarization or sectoral concentration. Yet they can coexist with significant wage premiums for individuals possessing these scarcer competencies. In lower-income settings, the same skills have neutral or even positive mediating effects, suggesting that demand for foundational digital services may still be expanding in these regions.

Perhaps most critically, the World Bank’s digital development framework aligns with the ILO’s finding that employers consistently seek a broad mix of complementary abilities. Online vacancy data from Brazil, Russia, South Africa, Egypt, and the UAE reveals that only 1–2% of job postings mention machine learning and AI skills. General computer skills appear in 5–10% of postings. Demand for AI-related skills is gradually increasing, particularly among high-skilled occupations, but socio-emotional skills remain essential complements.

In the UAE, machine learning and AI skills are linked to an 18% wage premium, but 88% of those roles also require social skills and 84% demand core cognitive skills. In South Africa, software-specific skills carry a 19% wage premium while being accompanied by social skills requirements in 70% of postings. The message is clear: technical AI skills in isolation are insufficient—the future workforce needs balanced, multidimensional competencies.

Algorithmic Management and AI Job Quality Concerns

While employment numbers capture one dimension of AI’s impact, the ILO emphasizes that the most profound effects may be felt not on net employment but on job quality. The rapid proliferation of AI-powered analytics and algorithms is fundamentally transforming workplace management in ways that demand scrutiny from labor regulators and worker organizations.

Businesses are increasingly deploying what the ILO terms “algorithmic management” for tasks spanning the entire employment lifecycle: hiring, performance monitoring, scheduling, and supervision. While promising increased efficiency and productivity, these AI-driven management systems carry significant risks for workers. Constant performance monitoring creates sustained pressure to meet algorithmically determined targets. Opaque decision-making processes leave workers unable to understand or challenge evaluations. And the automation of supervisory functions can diminish worker autonomy while intensifying workloads.

The ILO’s evidence suggests that workers subject to AI-driven monitoring may face chronic stress, struggle against inflexible algorithmic processes, and experience significant loss of control over their work conditions. These effects operate below the radar of traditional employment statistics—a company might maintain the same headcount while fundamentally degrading the quality and dignity of work through algorithmic oversight. For deeper analysis of how AI governance frameworks are evolving, explore our collection of regulatory and policy reports.

This dimension of AI impact requires policy responses that go beyond employment numbers. Labor standards need updating to address algorithmic transparency, worker data rights, and the limits of automated supervision. The G20 framework offers a platform for coordinating these regulatory approaches across major economies.

Data Labourers: The Hidden AI Workforce

Among the ILO paper’s most important contributions is its spotlight on the “data labourers” who form the invisible foundation of AI systems. Competitive pressures lead AI developers to present their products as fully autonomous solutions, obscuring the crucial human labor that makes these systems function. AI’s reliance on vast quantities of meticulously labelled, categorized, and annotated data necessitates a substantial hidden workforce.

These workers typically operate on microtask platforms or within artificial intelligence-business process outsourcing (AI-BPO) companies, performing repetitive tasks under precarious conditions. An ILO-Thunderbird survey conducted between 2022 and 2023 reveals troubling statistics about their working conditions: only 50% of BPO data workers reported access to paid holidays, 65% had sick leave, and 50% had access to maternity leave. Workers reported feeling trapped in short-term positions with limited opportunities for career advancement and intellectual fulfillment.

The geographic dimension compounds these concerns. Much of this data work is outsourced to developing economies, where lower labor costs and weaker institutional frameworks allow companies to implement working arrangements that would face regulatory barriers elsewhere. This includes excessive algorithmic management practices used to control worker behavior—creating an ironic situation where the technology built to manage other workers is itself sustained by a workforce subject to some of the most intensive surveillance and control.

The ILO calls for including data labourers in conversations about AI governance—a significant policy recommendation given that these workers are often rendered invisible by the very industry narratives their labor supports. Content moderators face additional risks of exposure to disturbing material, raising concerns about mental well-being that existing occupational health frameworks are poorly equipped to address.

AI Policy Implications for G20 Nations

The ILO’s analysis culminates in four interconnected policy pillars that G20 nations must pursue simultaneously to harness AI’s benefits while mitigating its risks. Each pillar addresses a distinct dimension of the challenge while reinforcing the others.

Investing in skills requires curricula reform from early childhood through higher education, integrating foundational digital literacy, critical thinking, creativity, complex problem-solving, and socio-emotional learning. For emerging markets and the African Union, the priority is closing foundational education gaps while building technical capacity. Upskilling programs for the existing workforce must focus on skill bundles rather than isolated technical AI competencies.

Strengthening social protection is essential given that traditional safety nets may struggle with increased job transitions. While global social protection coverage reached 52% of the population in 2023, only 19.1% of Africa’s population is covered. The ILO suggests AI could itself help extend protection by creating digital financial footprints that enable formalization of micro and small firms in the informal economy.

Fostering responsible AI innovation means creating environments that encourage beneficial AI development while establishing safeguards. Governments can use targeted R&D tax credits and regulatory frameworks to ensure AI augments human labor rather than constraining worker autonomy. The overarching objective is leveraging technological advances to improve both productivity and working conditions.

Enhancing international cooperation recognizes that disparities in AI exposure and digital capability demand collaborative global efforts. The United Nations AI Advisory Body and the G20 Framework Working Group provide platforms for exchanging policy experiences, coordinating regulatory approaches, and channeling financial and technical assistance to nations building their AI ecosystems.

The ILO concludes that the long-term consequences for jobs, inclusive growth, inequality, and social cohesion remain to be seen—necessitating careful monitoring and potentially new policy approaches to manage the transition. What is certain is that inaction will not preserve the status quo; it will merely cede control of the transition to market forces alone, amplifying existing inequalities rather than addressing them.