In the coming weeks, the European Commission is expected to table the Quality Jobs Act, new legislation that will shape how artificial intelligence (AI) and other technologies are used in workplaces across the region for years to come.
The timing makes the choices unusually consequential. Get it right, and Europe can set a global standard for technology that genuinely works for people. Get it wrong, and the result is fragmented rules that large companies have the resources to navigate, while smaller ones do not.
That tension ran through the Reshaping Work 2026 conference in Amsterdam, where Stefano Perego, Amazon’s Vice President of International Operations, joined Roxana Gheorghiu, Lawyer at the European Commission’s DG Employment; Stijn Broecke, Senior Economist leading the OECD’s Future of Work initiative; and Marcin Wolski, Economist at the European Investment Bank, for a panel called Making AI at Work, Work: Adoption and the Human Frontier.
The discussion surfaced five conditions that will determine whether Europe gets that balance right.
1. Invest in people alongside technology
The evidence is increasingly clear: technology alone concentrates gains at the top. Paired with skills investment, it spreads them.
A study by the Politecnico di Milano and Amazon found that 80% of companies using AI for demand forecasting reallocated staff to higher-value work, and 40% reported measurable improvement in digital skills—but only where training programmes were in place alongside the technology.
Amazon’s own experience bears this out. “The two things that matter the most,” Perego told the panel, “Are investing in the technology whilst we invest also in people. You cannot have technology thrive without having people thriving in it.”
Through Career Choice, the company pre-pays up to 100% of tuition across more than 100 programmes and 90 European institutions. Since 2014, more than 50,000 European employees have enrolled—that’s roughly one in five employees in the company’s European workforce—with €30 million committed for 2026 alone. The company also runs more than 20 apprenticeship schemes across Europe, with 540 apprentices currently on-scheme.
Christian Wirz joined Amazon as an IT apprentice in Germany in 2013. He is now a Cluster Manager for one of the largest AWS data centre clusters in Europe. “An apprenticeship is not a consolation prize for people who don’t attend university,” he said. “It’s a world-class pathway.”
OECD Senior Economist Stijn Broecke argues the debate itself needs to shift. “Many people are worried about automation and employment, and that leads to policy discussions focused on social protection, universal income, those kinds of things,” he said. “Whereas when you talk about augmentation — How do you get workers and companies to use AI in a way that creates more value? — that shifts the policy debate to investing in skills. And I think that’s really important.”
2. Keep humans in the loop
If the first condition is about investment, the second is about accountability. Where technology touches people’s work, a person must remain responsible for the decisions that matter.
There are three commitments Amazon applies across its operations. First, a human is always in command: no significant decision about someone’s job is taken without a person who is accountable for it. Second, expectations are fair: performance standards are based on what has proven achievable over time and take individual circumstances into account. Third, transparency: workers can ask how the systems that affect their work function, and employee representatives are informed when new systems are introduced.
In practice, this means AI informs decisions rather than making them. Across Amazon's fulfilment network, that principle plays out in concrete ways.
Perego pointed to maintenance operations, where tens of thousands of engineers keep the company’s infrastructure running. “What AI is helping us do is moving from a state of reactive maintenance to a state of prescriptive maintenance,” he said. “In almost real time, through sensors and through our global network, I know what happens in that specific machine in this moment in time, and I can send a person and support that person and intervene.” The engineer still makes the call; the technology tells them where to look.
The policy implication, Perego argued, is that the Quality Jobs Act should focus its obligations proportionately, on automated decisions that have a significant effect on someone’s job, not on every digital tool a manager uses.
“The technology must augment and assist human workers, not dictate what they do,” said Roxana Gheorghiu, Lawyer for the European Commission. “If we have a legal framework that is targeted, risk-based and proportionate—fully aligned with the GDPR and the AI Act—I think AI adoption will increase. We want to foster innovation. However, we cannot leave workers’ rights behind.”
3. Make AI inclusive by design
Inclusion might be the least-told part of the AI story. Across Amazon’s European network, nearly 5,000 employees have a disability. In Italy, the company has hired more than 150 deaf employees who use AI-enabled tools to receive tasks and communicate without intermediaries, giving them genuine workplace autonomy.
“Our ambition is that AI can help close the employment gap between people with and without disabilities,” Perego said. “What we see in our own operations suggests that’s achievable, but only when you design the tools inclusively from the start.”
The point extends beyond any single company. Legislation that only regulates against AI’s risks, without recognising and incentivising its inclusion potential, misses half the picture.
4. Ensure regulatory coherence across the Single Market
Europe already has substantial rules governing AI in the workplace. The GDPR regulates automated decision-making. The AI Act’s high-risk employment provisions take effect in December 2027. The OSH Framework Directive covers all workplace risks, including psychosocial risks. National labour laws add further layers.
The risk is fragmented regulation.
“The real divide isn’t a technology gap, it’s a capacity gap. Roughly half of large firms are adopting AI, but far fewer SMEs are keeping pace. A directive that adds complexity without closing that gap will widen it.”
New data from the European Investment Bank underscores the scale of the challenge. The EIB’s latest survey found that 67% of companies in the EU use generative AI, compared to 70% in the United States. But 13% plan to adopt in the next 12 months, which would bring the EU to a staggering 80% adoption rate.
According to EIB Economist Marcin Wolski, “There are still pockets of vulnerability, with the adoption rates among micro and smaller enterprises much below the larger peers. But when we asked why they’re not using AI, they seem to see no need for the technology. It's not that they lack skills or capacity to adopt. The task is to show smaller firms how AI can be useful.”
Both these barriers—regulatory complexity and a lack of perceived relevance—point to the same conclusion: smaller firms need simpler frameworks and stronger incentives to engage.
5. Build the transition infrastructure together
No single company, however large its investment, can build the infrastructure a labour market transition requires. That demands mutual recognition of AI-era skills across borders, co-investment by governments and employers, and genuine social dialogue.
Amazon is a signatory of the EU Pact for Skills and has committed €10 billion to expand and modernise our European fulfilment, logistics and technology network, an investment creating thousands of roles that interface with next-generation technology. Through Future Ready 2030, a $2.5 billion global commitment to prepare 50 million people for the future of work, free AI-skills training has already reached more than 2 million people in Europe.
But the transition cannot be built company by company. “We need policy frameworks that enable every employer to do the same,” Perego said. “The future of work is humans empowered by technology, and the choices we make now will determine how widely that future is shared.”