Generative AI Exposure in European Labour Market: Who Is Most Affected? (Bruegel 2025)

Why Generative AI Exposure Matters for European Workers

Generative artificial intelligence has moved from laboratory curiosity to workplace reality with unprecedented speed. Since the launch of ChatGPT in late 2022, organisations across Europe have rushed to integrate large language models, image generators, and AI coding assistants into their workflows. But which workers actually face the greatest disruption — or opportunity — from this technological shift? A rigorous 2025 Bruegel working paper provides the most comprehensive European-specific answer to date.

The study, authored by Laura Nurski and Nina Ruer, applies two distinct generative AI exposure measurement frameworks to the European Labour Force Survey — covering millions of workers across EU member states — to map precisely which demographic groups, occupations, and skill levels face the highest exposure. The findings challenge widespread assumptions about AI displacement and offer crucial evidence for policymakers, employers, and workers navigating the most significant labour market transformation in decades. For organisations seeking to understand how AI reshapes their industry, engaging with the full depth of this research through interactive analysis experiences can make complex findings accessible to every stakeholder.

How Bruegel Measures Generative AI Exposure

What distinguishes the Bruegel study from headline-grabbing predictions about AI replacing jobs is its methodological rigour. The researchers employ two complementary measurement frameworks to assess generative AI exposure across the European workforce. The first is a task-based approach, which maps the specific tasks within each occupation — such as text drafting, data analysis, customer communication, or report summarisation — to the demonstrated capabilities of generative AI systems. Occupations whose core tasks align closely with what GenAI can perform receive higher exposure scores.

The second framework uses an ability-based approach, assessing workers’ general cognitive abilities — verbal reasoning, numeracy, pattern recognition, and social interaction — against GenAI strengths. While both approaches identify broadly similar patterns, the researchers argue that task-based analysis produces more actionable insights for employers designing AI integration strategies and for policymakers assessing employment impacts.

Both exposure frameworks are merged with the European Labour Force Survey (LFS), which provides detailed microdata on workers’ demographics, education, occupation, sector, and employment status across all EU member states. This combination of AI-specific exposure measurement with Europe’s most comprehensive employment dataset produces a granular, continent-wide picture of where generative AI will make its greatest impact. Critically, the researchers emphasise that exposure does not equal displacement — it measures the potential for AI to affect job tasks, which can manifest as augmentation and productivity enhancement just as readily as substitution.

Generative AI Exposure by Gender: The Female Workforce at the Frontier

Perhaps the study’s most striking finding is the clear gender dimension of generative AI exposure. Across both measurement approaches and across European countries, women face higher GenAI exposure than men. This pattern reflects the occupational distribution of female employment across Europe: women are disproportionately represented in roles with high concentrations of language-intensive, administrative, and knowledge work tasks — precisely the activities where generative AI demonstrates its strongest capabilities.

Occupations in education administration, healthcare management, customer service, financial advisory, legal support, and professional services — all areas with strong female representation — score consistently high on GenAI exposure indices. In contrast, occupations with lower exposure tend to cluster in physical, manual, or outdoor work categories where male employment predominates, such as construction, manufacturing floor operations, and transportation.

This gendered exposure pattern demands careful policy attention. If generative AI adoption proceeds without targeted intervention, it could disproportionately transform or displace jobs held by women. However, the same exposure also represents an opportunity: women in high-exposure occupations may benefit most from AI-augmented productivity if organisations implement thoughtful integration strategies. The critical variable is whether employers use GenAI to enhance these roles or to eliminate them — and whether policy creates the right incentives for the augmentation pathway.

Education and Age Patterns in AI Exposure

The Bruegel analysis reveals two additional demographic patterns that challenge conventional wisdom about automation. First, highly educated workers face greater generative AI exposure than those with lower educational attainment. This reverses the traditional automation narrative, where routine manual and clerical tasks — typically held by workers with moderate education — were considered most vulnerable. Generative AI’s strength in language processing, reasoning, analysis, and content creation means it penetrates deep into knowledge work that previously required advanced degrees.

Professionals in law, consulting, finance, marketing, journalism, and academic research — roles typically requiring university education or higher — find their core tasks increasingly within GenAI’s capability range. This does not necessarily mean these jobs will disappear; rather, the nature of the work will transform. Tasks that once consumed hours of a professional’s time — drafting documents, conducting preliminary research, analysing datasets, generating reports — can now be accelerated dramatically with AI assistance, fundamentally reshaping how high-skill work is organised and valued.

Second, younger workers show higher generative AI exposure than their older colleagues. This partly reflects younger workers’ concentration in entry-level knowledge work roles with high task-based exposure. However, it also reflects generational differences in digital adoption: younger workers are more likely to encounter GenAI tools in their workflows and workplaces. The implications are double-edged. Younger workers face greater disruption risk if AI substitutes for their early-career tasks, but they also have longer career horizons to adapt, reskill, and leverage AI as a productivity multiplier.

Which European Occupations Face the Highest GenAI Exposure

The study’s occupational analysis identifies specific categories where generative AI exposure is highest. Knowledge-intensive service occupations — including business and administration professionals, legal professionals, finance analysts, and information technology specialists — consistently rank among the most exposed. These roles involve substantial amounts of text production, data interpretation, client communication, and analytical reasoning — all core GenAI capabilities.

Administrative and secretarial roles also show very high exposure, reflecting GenAI’s ability to automate correspondence, scheduling, document management, and data entry. Customer service and contact centre occupations face significant exposure as conversational AI systems become increasingly capable of handling complex customer interactions across multiple languages and channels.

At the other end of the spectrum, occupations involving physical dexterity, outdoor work, interpersonal care, or complex manual tasks show markedly lower GenAI exposure. Healthcare workers providing direct patient care, skilled tradespeople, agricultural workers, and protective service occupations maintain relatively low exposure scores — though this may change as AI capabilities extend beyond language and reasoning into robotics and physical automation. Understanding these patterns helps organisations identify which teams and roles require the most urgent AI readiness investments and strategic planning resources.

The Equalising Effect: How AI Boosts Less-Experienced Workers

One of the study’s most hopeful findings concerns within-occupation heterogeneity — the question of who benefits most from GenAI assistance among workers doing the same type of job. Reviewing evidence from multiple experimental and observational studies, the Bruegel researchers find a consistent pattern: less-experienced and less-skilled workers within an occupation gain the largest productivity improvements from generative AI tools.

This pattern has been documented across diverse contexts. In customer service, agents with less experience show larger performance gains when assisted by AI suggestions. In software development, junior programmers benefit more from AI code completion tools than senior developers. In writing and analysis tasks, workers with weaker baseline skills see their output quality converge toward the level of more experienced colleagues when supported by GenAI.

This equalising dynamic suggests that generative AI could serve as a powerful tool for reducing within-occupation inequality — democratising access to expertise and accelerating skill development. However, this outcome is not automatic. It depends critically on how organisations deploy AI tools, whether they share productivity gains with workers, and whether policy frameworks incentivise augmentation over substitution. Without deliberate design, the same tools could instead be used to deskill jobs, compress wages, or justify reducing headcount in roles where AI narrows the performance gap between experienced and entry-level workers.

Task-Based vs Ability-Based AI Exposure Analysis

A significant methodological contribution of the Bruegel study is its systematic comparison of task-based and ability-based exposure measurement. While both approaches identify broadly consistent demographic patterns — women, the highly educated, and younger workers as most exposed — they differ in important ways that affect their practical utility.

The task-based approach maps specific workplace activities to GenAI capabilities. This granularity makes it directly actionable: employers can identify which tasks within a role are candidates for AI augmentation, redesign job descriptions accordingly, and target training to the specific skills workers need to collaborate effectively with AI systems. The task-based lens also supports more precise assessment of employment effects, distinguishing between roles where AI will handle a small share of tasks (modest impact) versus those where it can perform most core activities (transformative impact).

The ability-based approach operates at a higher level of abstraction, mapping general cognitive and social abilities to AI strengths. While useful for broad workforce planning and educational policy, it is less immediately actionable for individual organisations. The Bruegel researchers conclude that the task-based framework offers greater practical value for both workplace integration strategies and employment policy design, recommending that future monitoring of AI’s labour market impact adopt task-level granularity wherever possible.

Labour Supply Policies: Training, Safety Nets and Reskilling

Based on their findings, the Bruegel authors propose a comprehensive dual policy framework addressing both sides of the labour market equation. On the supply side, the recommendations focus on preparing workers for a GenAI-transformed workplace. First, targeted reskilling and upskilling programmes must reach the most exposed groups — particularly women in administrative and knowledge work roles, highly educated professionals whose tasks are being automated, and younger workers entering a rapidly changing job market.

These training investments should go beyond basic digital literacy to address the specific competencies needed for human-AI collaboration: prompt engineering, AI output evaluation, ethical AI use, and the higher-order judgement skills that remain distinctly human advantages. The OECD’s framework for AI and employment provides useful guidance on structuring such programmes at national and European levels.

Second, social safety nets must be strengthened to support workers during transitions. This includes adequate unemployment insurance, portable benefits that follow workers across jobs and platforms, and transitional income support for those undertaking substantial retraining. Third, active labour market policies should provide personalised career guidance that accounts for AI exposure patterns — helping workers in high-exposure occupations identify transition pathways to roles that leverage their existing expertise while building AI-complementary skills.

Labour Demand Policies: Job Redesign and Organisational Agility

The demand-side policy recommendations may be even more consequential. The Bruegel authors argue that the outcome of generative AI exposure depends fundamentally on how employers choose to integrate these technologies. Policy must therefore shape employer behaviour, not just worker preparedness. The central recommendation is to incentivise job redesign that uses AI to enhance role quality rather than simply eliminate positions.

This means creating frameworks that encourage organisations to: redistribute AI-generated productivity gains through higher wages or reduced working hours; redesign roles to combine AI-automated tasks with higher-value human activities; and invest in organisational processes that support effective human-AI collaboration. Tax incentives for firms that demonstrate job quality improvement alongside AI adoption, public procurement standards that reward augmentation-oriented AI strategies, and sectoral collective bargaining that addresses AI integration terms could all contribute to steering outcomes in the right direction.

Organisational agility — the ability of firms to rapidly adapt their structures, processes, and workforce strategies to technological change — emerges as a critical competence. Firms that adopt rigid, substitution-focused approaches to AI will likely destroy jobs and degrade remaining roles. Those that embrace agile, augmentation-focused strategies can capture productivity gains while creating better jobs. Policy can encourage the latter through management training programmes, best-practice sharing platforms, and regulatory frameworks that make the augmentation pathway the path of least resistance. For leaders managing this transition, access to interactive analyses of workforce research can accelerate understanding and decision-making across all levels of the organisation.

Navigating the Jagged Technological Frontier

The Bruegel study concludes with a call to monitor and research what they term the “jagged technological frontier” — the uneven, unpredictable boundary between tasks that generative AI performs well and those where it falls short. This jaggedness means that exposure patterns will continue to shift as AI capabilities evolve, making ongoing monitoring essential for effective policy.

National statistics agencies and labour ministries across Europe must develop systematic approaches to tracking GenAI adoption and its employment effects. This includes integrating AI exposure indicators into regular labour force surveys, establishing employer-side surveys on AI implementation strategies, and creating early warning systems for sectors and regions facing rapid transformation. Research funding should support longitudinal studies that track the actual — not just predicted — impacts of GenAI on job quality, wages, career progression, and inequality.

The Bruegel working paper provides an essential evidence base for this effort. Its clear finding that generative AI exposure follows demographic lines — disproportionately affecting women, the highly educated, and younger workers — demands that Europe’s response be equally targeted. The combination of supply-side training investments and demand-side job redesign incentives, guided by ongoing monitoring and grounded in rigorous research, offers the best path toward a future where generative AI serves as a tool for shared prosperity rather than a driver of deeper inequality. For professionals and organisations committed to navigating this transition wisely, deep engagement with the underlying evidence is not optional — it is essential.