AI Redefines Work: Autonomy or Fragmentation?

The Profound Impact of AI on Organizations

Artificial Intelligence (AI) is not just changing the tools we use daily. It is redefining fundamental concepts such as autonomy, expertise, and collaboration within organizations. While companies have made progress in structuring governance around AI models, data, and compliance, this is no longer sufficient. The real challenge lies in integrating these technologies into the daily lives of teams and the impact they have on employees. In this regard, many organizations are already significantly lagging behind.

A New Divide: Act or Suffer

Generative AI introduces a new divide within companies, subtler than the digital divide of the past. It no longer simply distinguishes those who have access to technology from those who do not, but separates those who make decisions with the help of AI from those who merely follow its recommendations. Maintaining control over decisions and the meaning of one’s work is essential for those who take action. Conversely, suffering from AI means passively accepting its results, often perceived as quick and reliable, without any effort for analysis.

This boundary is not a matter of skill but of intention. It runs through every team, every profession, every day. Generative AI, described as a "general-purpose technology" by innovation economists, reconfigures work and the balances of expertise and agency, much like the printing press or electricity.

Thought in the Face of AI's Speed

Action requires reflection, and this reflection demands time that AI tends to reduce. This tension is central to the debate on AI, although it is rarely articulated explicitly. When AI assistants produce in seconds what previously took hours of work, two phenomena occur simultaneously: a time gain and a decrease in analytical effort. As delegation increases, skills erode.

Preliminary research from MIT Media Lab, published in 2025, shows a decline in brain connectivity among heavy users of language models for writing. Work psychologists also observe a decrease in the sense of personal effectiveness when expert actions disappear.

This phenomenon is not new. As early as the 16th century, La Boétie questioned the voluntary submission of free men, not through force, but through the comfort of no longer having to decide. In the 20th century, Erich Fromm spoke of the fear of freedom, to which delegation to AI offers relief. However, this relief carries strategic risks.

To avoid suffering from AI, organizations must preserve spaces for reflection, even if they sometimes seem unnecessary, as this is where judgment is formed.

The Silent Fragmentation of the Collective

AI also alters the collective within organizations. It responds quickly, without mood or agenda, becoming the natural interlocutor for questions once posed to colleagues or managers. This shift may not be immediately noticeable in production indicators, but it affects the living memory of the organization in the long term.

With a universal assistant at hand, each employee risks becoming a brilliant soloist, equipped with the same cognitive prosthesis, working in parallel without interaction. This scenario of omnipotent individualism perceives dependence on the collective as a hindrance and deliberation as a cost.

Yet, true expertise, which supports an organization in difficult times, lies in the collective's ability to debate, arbitrate, and transmit knowledge. The work of Amy Edmondson and Google's Project Aristotle highlights the importance of psychological safety, which allows for the expression of misunderstandings, mistakes, or disagreements. This is currently the most exposed factor and is rarely measured in AI projects.

AI: A Lever for Engagement or an Accelerator of Withdrawal

Leaders face a crucial choice that is not technological but anthropological: what will be the future of work and workers in this transformation?

AI can be a powerful lever for engagement, freeing up time for non-delegable tasks: making decisions in uncertainty, nurturing relationships, transmitting skills, debating directions. It can restore to managers the leeway lost under operational pressure. It can elevate the level of collective performance, provided that the freed-up time is reinvested in what strengthens the collective, rather than simply increasing the pace.

Conversely, introducing AI as a mere productivity gain, disconnected from work, without recognition of roles, and without dialogue with teams, can accelerate withdrawal into oneself. Uses develop on the margins of the company's functioning, managers improvise, and the Executive Committee discovers the effects only once they are established. European regulators set a deadline with the AI Act, applicable to high-risk systems starting August 2, 2026, but no regulation will replace an informed internal governance decision.

Governing the Human Impacts of AI

The governance requirements for AI models, such as traceability, scope, reversibility, and measurement, have their equivalents on the team side.

• It is crucial to trace, job by job, what AI takes on and what it modifies in the expertise chain.
• Explicitly decide what will not be delegated, in consultation with the relevant employees, as the decision is not solely technical.
• Preserve the reversibility of the organization by maintaining spaces for transmission, collective rituals, and areas for learning through practice.
• Measure, in the long term, the impact on engagement, sense of purpose, and the quality of the collective.

These requirements do not fall solely on HR, but on the Executive Committee, just like cybersecurity or compliance, as they touch on an essential intangible asset: the collective capacity to think together.

The question for leaders is no longer whether AI will be integrated into the company; it already is, whether they have decided so or not. The real question is to determine what we want to continue doing ourselves because we deem it important. It is this choice that distinguishes a proactive organization from a passive one.