by Ummuhan Bardak, Senior Human Capital Development Expert, ETF
Introduction
Artificial intelligence (AI) has been one of the greatest disruptions in labour markets, transforming tasks and changing how work is performed and valued. From recruitment software to automated scheduling, from generative tools supporting knowledge work to algorithmic systems coordinating tasks, AI's presence is no longer a futuristic scenario. It is reshaping society at a speed that public institutions struggle to match. Regulations arrive late, training systems adapt slowly, and policymakers are left navigating a landscape that changes faster than any framework can capture.
So what should governments, workers and educators do? This is precisely why understanding AI's impact across sectors, occupations, working conditions and different social groups and countries is so urgent. Only when we know what we are dealing with, can we make the most of the opportunities AI offers while taking action at the same time to reduce its risks.
The European Training Foundation’s latest analysis of the impact of AI on labour markets brings clarity to a debate often shaped by speculation. Drawing on international evidence and recent empirical studies, the report shows that AI’s effects are often uneven and shaped by policy choices. What emerges is neither a story of mass job destruction nor one of automatic progress, but a more complex picture of work in transition.
An evidence‑based overview of how AI is reshaping work, the report provides some key takeaways for experts, practitioners and decision-makers in the ETF’s partner countries who will shape the collective future of their countries.
Jobs are changing fast but not disappearing
The key effect identified so far has been the transformation of existing jobs that undergo significant task changes, but only a limited number of roles can be fully automated. Most occupations combine tasks that can be automated, tasks that are reshaped by AI support, and tasks that remain firmly human.
The result is a reshaping of how work is done, not a sudden disappearance of work itself. At the same time, the report flags a structural risk that deserves close attention: Entry‑level jobs are more exposed than senior roles. Many routine cognitive tasks traditionally assigned to junior staff are now automated or assisted by AI systems. This weakens career ladders and raises longer‑term concerns for gaining seniority and professional progression.
Uneven AI effects by age, gender, education and occupation
AI does not affect all workers in the same way. Highly educated and digitally skilled workers tend to benefit most, often using AI as a productivity‑enhancing tool that supports learning and autonomy. On the other hand, lower-skilled workers experience AI less as an assistant and more as a source of pressure.
The latter is linked to the growing use of algorithmic management. These systems automate managerial functions such as task allocation or performance monitoring. Once associated mainly with platform work, they are now spreading into logistics, manufacturing, services and office‑based environments. Their impact depends less on the technology itself than on how organisations choose to deploy it.
In some contexts, algorithmic tools improve coordination or support safer working practices. In others, they intensify work, reduce discretion and increase workplace monitoring and surveillance (so-called ‘datafication’ of the workplace). This expansion of workplace data collection raises questions about privacy and data protection of workers, transparency and the balance of power at work.
Job quality is the key battleground
The most immediate effects of AI are visible in job quality, not employment numbers. The report documents a wide range of outcomes. Some workers experience more engaging tasks and clearer access to learning opportunities. Others face higher work intensity, limited scope to use their skills, or increased psychosocial strain. Low‑skilled and routine jobs are more likely to experience work intensification without corresponding gains in autonomy or pay. High‑skilled roles often benefit from AI‑driven support, yet even here the pace of work can increase. These mixed effects underline the importance of organisational practices and labour institutions in shaping outcomes.
The negative outcomes for job quality often stem from organisational factors and management choices, rather than the technology itself. This, in turn, is affected by the moderating role (or lack thereof) of institutional and regulatory frameworks, as well as by the organisational structures of firms and work culture. As AI systems tend to replicate existing power dynamics in organisations, their impact on worker well-being depends on how it is implemented in the workplace and whether workers have any say in it. Consequently, disadvantaged groups may suffer more from the negative effects of AI, risking further labour market polarisation across socio-economic groups.
Inequality risks are structural
AI tends to amplify existing labour market inequalities by favouring privileged groups over disadvantaged ones and it may compromise the access of disadvantaged groups to decent jobs. The power imbalance in AI development and implementation is pervasive across gender, race
and socio-economic background, and it is often highly educated managers and technology developers who decide the features of AI systems from their own perspectives and interests.
Education remains the strongest predictor of who benefits. Young male workers with higher education gain the most from AI. The level of impact escalates with income, creating several new opportunities for highly educated and high-income professionals who tend to be prime-age white male workers. Workers with limited digital skills or lower incomes face higher risks of job degradation. Women are overrepresented in clerical and administrative roles with higher exposure to automation and remain underrepresented in AI development and advanced digital occupations.
AI systems increasingly influence who gets seen, selected, promoted, or dismissed in the labour market and the report provides evidence for persistent bias in AI systems used for recruitment, evaluation and promotion. When trained on historical data, these tools can reproduce existing patterns of discrimination. Without deliberate corrective action, the digital divide risks hardening into an AI divide.
AI may widen global inequalities
Developing countries start the AI age with disadvantages, often exacerbated by the global digital divide, leading to highly uneven AI adoption across countries. Advanced economies, supported by stronger digital infrastructure and broader skills bases, are better positioned to benefit from AI‑related productivity gains. Many developing countries face a different starting point, with limited access to digital technologies and fewer opportunities for AI‑related skills development.
At the same time, millions of workers in lower‑income countries are integrated into the AI value chain through low‑paid micro‑task work such as data annotation and content moderation. These jobs are essential to AI development but often involve weak protections and intensive monitoring. Without policy intervention, AI risks widening global inequalities.
Policy choices will shape outcomes
AI can drive innovation, increase productivity, and open new frontiers for economic growth. But it can also reinforce exclusion, inequality and insecurity if left unchecked. Ultimately AI’s impact is not predetermined. The challenge for policymakers is to actively shape the conditions of AI deployment so that it supports – not undermines – human dignity, equality and opportunity. Countries with strong labour protections, effective social dialogue and forward‑looking skills policies are better equipped to steer AI towards job upgrading rather than job erosion. Limited institutional and regulatory frameworks in developing countries create a higher risk of AI having a negative impact.
Regulation has an important role to play. Recent EU frameworks set safeguards for high‑risk AI systems used in employment and education. Yet regulation on its own is insufficient. The report points to the need for sustained investment in digital and AI literacy, targeted upskilling and reskilling, and social protection systems that can adapt to changing forms of work. Policy choices, governance choices and ethical choices will determine the outcome. And the ETF's role is to communicate this message and evidence so that policy decisions are taken wisely.
Please log in or sign up to comment.