In February 2026, two major events placed AI at the center of the HR debate: the AI Action Summit organized by the Ministry of Labor and Solidarity and LinkedIn France’s Talent Connect Session. The conclusion from experts and executives present was unequivocal: AI does not raise everyone’s skills equally; it creates gaps. In France, it is estimated that 20% of jobs are highly exposed to AI and only weakly protected from automation^[1]. And while 38% of French workers say that AI makes them more efficient^[2] ,the reality in the workplace is more nuanced: AI adoption remains scattered, cognitive load is increasing, and some employees prefer to conceal their practices (what is known as shadow AI).

How can organizations deploy AI in a way that strengthens teams rather than weakening them?

We identified three actions to take:

• Establish an ethical framework before deploying tools;
• Anchor AI in actual work rather than prescribed work;
• Involve employee representatives from the start rather than at the end of the process.

Establish an ethical framework before deploying tools

The temptation to deploy AI tools quickly and train teams afterward is strong. This is often where resistance begins. What works is the opposite approach: first establish a framework of trust and make it meaningful, then train people. Defining this framework of trust cannot happen without involving IT teams.

For example, SAGE structured its approach around a clear principle: address AI from an ethical perspective before training employees on tools. The AI Academy, created two years ago, systematically begins its program with ethics training. “AI champions” then reinforce internal awareness, while “AI snacks” — informal moments for discussions and exchanges about AI — keep AI alive in teams’ day-to-day. Before each deployment, one question is asked: “Why are we using AI for this specific purpose, and to what end?” This approach applies to all employees, including those who have been with the company for twenty or even forty years.

Schneider Electric made a similar choice. As early as 2021, the group centralized AI solution procurement within an AI Hub, supported by a trust charter and an ethics committee bringing together data scientists, legal experts, and business leaders who review use cases almost in real time. This foundation enabled the company to roll out large-scale training programs in partnership with employee representatives.

With 200,000 employees worldwide, EDF validated five non-negotiable principles for any deployment of generative AI at the executive committee level: secure, user-supportive, non-discriminatory, environmentally responsible, and trustworthy. This framework acts as a systematic filter before launching any new initiative, preventing the uncontrolled proliferation of tools and the resistance that often follows.

Anchor AI in actual work rather than prescribed work ^[3]

“AI is based on prescribed work, not actual work,” reminds Emmanuelle Léon, Associate Professor of Work and Human Relations and Scientific Director of the “Reinventing Work” Chair at ESCP. This is the most common trap: implementing tools without understanding how work is truly carried out. Blind spots accumulate, distrust grows, and divisions deepen.

The experience of administrative assistants during the rise of information technology is particularly revealing in this regard. Conventional wisdom held that it would be easy to do without them because digital tools would absorb their work. What was underestimated, however, was that they were often the living memory of organizations: they knew whom to call during a crisis, how to resolve complex situations, and which procedures could be bypassed to move a case forward effectively. No software could replace that role. With AI, the risk is identical: deploying tools based on visible work — the work that is written, formalized, and prescribed — while ignoring the invisible aspects of work that are equally, if not more, important.

Involve employee representatives from the outset rather than at the end

Organizations that neglect social dialogue around AI often face costly resistance during implementation. Isabelle Quainon, Chief Human Resources Officer at Veolia, expresses this clearly: “The worst thing that could happen to us is for AI to become an inescapable fate imposed on workers.” Avoiding that fate requires governance built with employee representatives rather than imposed upon them.

EDF established a three-level social dialogue structure: a central works council, a France Group Committee, and local works councils. The company also launched workforce planning initiatives specific to AI and developed an internal communication toolkit to support managers in discussions with their teams, transforming every conversation about AI into an opportunity for co-creation rather than a top-down announcement.

SAGE, for its part, made social dialogue a prerequisite for deploying every new AI initiative. The company notably developed a talent marketplace that promotes internal mobility without criteria related to background or academic qualifications — a concrete way of placing equity at the heart of transformation.

AI will not replace teams — it will test their strength

The deployment of AI within organizations is a transformation topic that must be driven by executive leadership and implemented jointly by IT, business teams, Learning functions, and HR departments. Organizations are progressing methodically: differentiated learning pathways based on expertise, governance frameworks established before tool deployment, and active social dialogue that gives meaning to every initiative.

What these experiences share is a conviction summarized by Karin Kimbrough, Chief Economist at LinkedIn: the objective is not simply to deploy tools but to “build a resilient talent pool with diverse skills and high adaptability.”

Because training employees on AI means engaging in a much deeper transformation than simply teaching them how to use tools. It encompasses decision-making, the role of managers, and the organization of collaboration between humans and machines. The organizations advancing most rapidly are those that understand that AI first challenges their operating model — how teams collaborate, how decisions are made, and how value is collectively created.

The real competitiveness question is therefore not: “Which tools should we adopt?” but rather: “How do we transform organizations to integrate these new tools sustainably?”

This is precisely where leaders have a decisive role to play: not as sponsors of an AI project, but as architects of a comprehensive transformation that leaves no one behind.

References

[1] Emmanuelle Léon, Associate Professor of Work and Human Relations & Scientific Director of the “Reinventing Work” Chair, ESCP

[2] LinkedIn France Talent Connect Session, February 2026

[3] Definition of prescribed work (French National Agency for the Improvement of Working Conditions): prescribed work is the work requested by management, sponsors, or leadership. It is often formalized through procedures to follow, job descriptions, and annual objectives, and is associated with expected outcomes.