By Ana Catarina de Alencar, international lawyer and ethicist, specialising in AI governance and technology regulation at the AI Collective.
This blog is produced as a part of the Reshaping Work 2025 Online Conference. The opinions and views expressed in this publication are those of the author, who presented his insights at the Online Conference. They do not purport to reflect the opinions or views of Reshaping Work or organisations that have supported the programme.
The rise of Artificial Intelligence (AI) agents is not just another chapter in the history of automation. It may be the turning point that redefines what human labor means. Unlike earlier technologies that replaced mechanical or repetitive tasks, AI agents integrate across platforms, execute complex workflows end-to-end, and increasingly act with autonomy. They don’t just assist us; they substitute, reorganize, and sometimes outpace us.
The effects are already visible. A Stanford study shows that employment among young workers in AI-exposed occupations has dropped by about 13% compared to peers in less exposed jobs. Companies like Klarna, British Telecom, and Shopify are restructuring their workforces, with AI assistants now handling vast portions of customer service and operational tasks. While some professionals benefit (workers with AI-related skills are commanding wage premiums of over 50%) many others face displacement or deskilling. The result is a polarization of opportunities that risks widening inequality and undermining social mobility.
Why AI Agents Are Different
Three features make AI agents distinct from earlier waves of automation: autonomy, integration, and adaptation.
- Autonomy: AI agents can interpret goals and act on them independently. Imagine an agent that notices your car needs repair, schedules the appointment, negotiates the cost, and completes the payment: all without your direct involvement.
- Integration: Agents connect multiple platforms into a seamless flow. In the workplace, they might draft reports from Slack, update a project board, schedule meetings, and prepare a presentation in one continuous chain.
- Adaptation: These systems learn over time, tailoring outputs to user preferences. In finance, emotional AI is even being tested to recommend decisions based on a user’s mood or stress level.
This shift transforms agents from tools into what some call “superagency”: systems that share decision-making capacity with humans and shape the architecture of daily life.
Who Loses Out?
The first wave of disruption is hitting junior professionals and entry-level white-collar workers. These jobs have long been gateways into careers, offering opportunities to acquire tacit knowledge through practice. If agents increasingly take over these functions, younger generations may never develop the same expertise. Researchers warn of the risk of cognitive atrophy: a future in which workers skip the foundational learning curve because machines do it for them.
The consequences are not only economic. Surveys show widespread anxiety: more than 70% of workers fear being replaced by AI, while many simultaneously feel pressured to adopt it to remain competitive. This paradox creates deep psychological strain, especially for younger professionals who see the ladder of progression pulled away just as they try to climb it.
Towards a New Social Contract
If left unchecked, AI agents could hollow out the entry-level layer of the workforce and accelerate inequality. But this outcome is not inevitable. Societies have choices, and they can build a new social contract that redefines the balance between human potential and machine efficiency.
What might this look like?
- Rethinking junior roles: Instead of treating young workers as low-cost labor, companies should cultivate their strengths in judgment, ethics, creativity, and human relationships: skills that machines cannot replicate.
- Education reform: Schools and universities must move beyond rote memorization and technical drills, preparing graduates to collaborate with intelligent systems. Programs like Singapore’s SkillsFuture show how continuous learning can equip people for multiple reinventions throughout their careers.
- Corrective mechanisms: In the future, governments may need to consider quotas or preferential hiring schemes for young workers in highly automated industries, ensuring access to learning opportunities that build long-term resilience.
- Redistributive measures: Experiments with universal basic income and wage insurance suggest ways to cushion transitions, giving workers time to retrain without falling into poverty.
- Responsible adoption: Regulation, such as the EU AI Act, emphasizes meaningful human oversight in high-risk domains. Extending this principle could preserve professional spaces for human judgment in finance, law, and customer relations.
A Political and Ethical Choice
Ultimately, the speed and scope of AI adoption are not just technical questions. They are political and philosophical choices. Do we want to automate everything that can be automated? At what cost to social cohesion, dignity, and opportunity?
The companies that manage to be efficient while still putting people first will not only stay strong in the market, also helping to create a more stable society for everyone. Governments that pace automation and protect vulnerable groups will foster trust and legitimacy. And workers, especially the youngest, must be given the chance to develop the human capabilities that machines cannot replace.
The rise of AI agents is more than a technological revolution. It is a turning point in the story of work. Whether it becomes a story of exclusion or empowerment depends on the choices we make today.