AI at work: productivity gains for whom?

AI adoption at work has doubled in two years. How Europe distributes the gains is still a choice it can make.

Young people entering Europe’s labour market are being asked to prepare for a future that even the people building artificial intelligence (AI) increasingly admit they cannot confidently predict.

In the space of a few years, the public message from the frontier of AI has ranged from extraordinary abundance and dramatically higher productivity to mass disruption of white-collar work and existential risk. Now some of the industry’s fiercest rivals are converging on caution. On 12 September 2026, Anthropic’s Dario Amodei called for the pace of frontier AI development to slow so that safety measures can catch up, a position now backed by Sam Altman and Elon Musk.

The uncertainty is not a reason for policy paralysis. It is a reason to be much more precise about what we do know.

Young people across OECD economies are already finding it harder to enter the labour market, particularly graduates. But the OECD is careful about causality: in its 2026 assessment, it notes that the deterioration predates ChatGPT and other large language models, and that current evidence does not establish AI as the cause.

AI is not necessarily creating Europe’s youth employment problem, but it is arriving at an unusually difficult moment for young workers, while potentially changing the very tasks through which they traditionally enter professional life.

Meanwhile, AI adoption is moving much faster than clarity about skills, rights or career pathways. The European Central Bank’s (ECB) survey of euro-area workers, published on 26 August 2026, finds that AI use at work doubled from 26% in 2024 to 52% in 2026. This is evidence of rapid adoption, not yet of economy-wide productivity gains.  What is  clear is that is becoming part of everyday working life before its longer-term consequences are settled.

We are automating the training ground

Generative AI is particularly capable at precisely the work through which young professionals traditionally learn their professions: drafting emails, taking minutes, summarising documents, transcribing and producing preliminary analysis.

Digital Futures Lab describes this as a potential new entry barrier. These tasks may appear repetitive to someone established in a profession, but they allow new workers to learn how organisations function, what good work looks like and how professional judgement is developed.

A task can be automatable and still be valuable because of what a human learns by doing it. Europe therefore needs to look beyond forecasts of occupations created or destroyed. A profession can survive while the route into it becomes narrower.

Workers themselves seem to understand the tension. Anthropic’s survey of more than 81,000 AI users, published in April 2026, found that early-career respondents were more concerned about displacement and that people reporting some of the largest productivity gains could also be among those most anxious about their future. The contradiction is understandable: workers can recognise that AI makes them more productive and simultaneously understand why employers may eventually decide they need fewer of them.

Firms can reverse course, workers cannot

Firms can experiment with organisational models. Workers experience those experiments as lives.

Klarna offers a useful warning against assuming that substitution is straightforward. The Swedish fintech became an emblem of AI-led restructuring after saying its customer-service chatbot could perform work equivalent to hundreds of employees. It later moderated that approach, restoring greater human capacity as it confronted the limits of automation in areas requiring nuance, judgement and service quality.

Klarna is not alone. Ford has reportedly rehired and promoted more than 350 experienced engineers after automated quality systems missed what veteran staff caught. Forrester’s Future of Work 2026 research finds that 55% of employers regret cutting jobs for AI and expects half of those layoffs to be reversed.

The lesson here is that companies are still discovering the frontier between machine efficiency and human judgement. And mistakes are asymmetric. A company can change strategy, reopen vacancies and redesign a service. A young worker who did not receive the role through which expertise was once accumulated cannot simply rewind that part of a career.

Productivity for whom?

None of this is an argument for Europe to slow AI adoption. Europe needs productivity. Its societies are ageing, labour shortages are increasingly structural, and public institutions and firms are under pressure to do more with fewer people.

But the familiar debate between ‘innovation’ and ‘regulation’ is too narrow. The more important questions are: what form of innovation, governed by whom, and producing gains for whom?

Productivity is not the same as worker agency or prosperity. Who controls the technology, who determines how it enters the workplace and who captures the resulting gains are separate questions.

European policymakers should therefore stop asking only how quickly firms adopt AI or how many jobs are gained or lost. They should also ask whose labour becomes recognised, valued or invisible as AI enters production, whether workers gain from the additional productivity they create and whether new routes into expertise are being built as old ones disappear.

Information is not bargaining power

Europe has an underused advantage here. Its distinctive asset is not simply that it regulates technology more heavily. It has institutions through which employers, workers and governments can negotiate how economic change is absorbed.

AI is already entering collective bargaining and social-partner negotiations across Europe. Eurofound, in research published in September 2025, documents agreements and initiatives addressing deployment, transparency, training and working conditions, although coverage remains uneven.

The Commission’s forthcoming Quality Jobs Act is therefore a significant opportunity. Its current consultation explicitly covers algorithmic management and AI at work, transparency, human-centred decision-making, skills and social dialogue. But transparency is not bargaining power.

In countries where institutions of co-determination and collective bargaining already exist, they should be used not simply to consult workers about technological change but to give them real influence over how it is introduced and how its gains are distributed. Workers should have a meaningful right to negotiate over AI systems that materially direct, evaluate, allocate or reorganise their work. Employers should be expected to explain not only what a system does but why it is being introduced and what it is intended to change. Workers should have a voice before substantial workplace redesign occurs, rather than simply being informed afterwards. Public administrations, as major employers and increasingly important adopters of AI, should lead by example.

The direction of technological change is not simply something societies must adapt to after the fact. It can be shaped.

The social contract is already at stake

For several decades, the social contract has rested, imperfectly but importantly, on the expectation that education, work and contribution buy some degree of economic security and voice. If younger people conclude that they can do everything society asked of them and still find the ladder pulled upwards, their frustration will not remain confined to the labour market.

A generation that loses faith in economic mobility can eventually lose faith in the institutions that promised it.

This does not make AI the cause of Europe’s wider economic or democratic malaise. It makes the distribution of its gains a test of whether European institutions can still manage economic transformation fairly.

Europe’s choice is therefore not between innovation and protection. It is whether to pursue AI adoption narrowly as a race for efficiency, or use it to renew a social bargain in which firms become more productive, public institutions more capable and workers retain a meaningful claim to the prosperity they help create.

Young people are not simply resisting technology. They are asking whether the institutions governing the transition are on their side.

Europe may struggle to match the United States or China on the scale of frontier AI investment. But it possesses something that will matter just as much as adoption accelerates: institutions capable of negotiating how technological change enters society rather than merely absorbing its consequences.

Those institutions cannot remain spectators. Europe does not need to protect every existing task. It needs to protect people’s ability to build a future from their work.

The views expressed in this #CriticalThinking article reflect those of the author(s) and not of Friends of Europe.