The diagnosis

On February 17, 2026, Benoît Serre was the guest of the sixth episode of the podcast "Human Footprint," the first French podcast dedicated to mental health at work. The title is stark: "Deterioration of Mental Health at Work." Not "evolution." Not "transformation." Deterioration.

Where the subject could devolve into catastrophism, Serre offers something more useful: he diagnoses the situation, identifies structural causes, and proposes a managerial framework for action.

Mental health at work has deteriorated in most sectors and geographies over the past five years, particularly among young workers (< 35 years), women, and screen-based jobs (executives, support functions, services).

The causes are not a mystery: work intensification, loss of autonomy in "how" tasks are performed, proliferation of processes, persistent post-COVID isolation, blurring of boundaries between professional and personal life, and more recently, anxiety related to AI itself (fear of replacement, stress of reskilling, cognitive overload).

The key quote

"The manager is the primary guardian of their team's mental health, provided they are given the means, time, and trust to do so."

Three words, three cumulative conditions. Without all three, the manager becomes a factor in the deterioration of mental health. This is likely the most important observation of this episode.

The direct link with AI

This is where the conversation becomes particularly useful for understanding 2026. AI, in its negative form, can exacerbate each of the three factors:

Exacerbation 1 : loss of autonomy. If AI makes decisions for the manager (ticket prioritization, HR recommendations, talent scoring), the manager loses their ability to arbitrate. They become a validator of AI outputs. However, autonomy is a key protective factor for mental health.

Exacerbation 2 : intensification. If AI frees up two hours but we allocate three more hours of work, the feeling becomes: "we're always asking for more." This is the phenomenon of "produced surplus": productivity increases, well-being decreases.

Exacerbation 3 : isolation. If AI answers questions on behalf of a colleague or a manager, it reduces quality human interactions, which are the primary protection against workplace anxiety.

But AI can also relieve

It would be unfair to only see the risks. AI can practically relieve managers by:

• Removing administrative tasks and meaningless reporting.
• Early detection of weak signals (absenteeism, decreased engagement, email overload) that can trigger earlier human managerial intervention.
• Offering a "first level" of listening (support chatbots, resources available 24/7) that eases non-critical cases and frees up the manager for cases that truly require their presence.

Serre is nuanced: the tool is not neutral. Its effect depends on the framework in which it is deployed.

Four managerial principles to keep

1. Protect human arbitration on high-impact decisions

Recruitment, evaluation, mobility, separation: AI can assist in preparation, but never decide on its own.

2. Ensure a portion of autonomy in "how" tasks are done

If everything is dictated by AI, work loses its meaning. There must be an explicit margin for maneuvering.

3. Convert saved time into presence, not additional workload

If AI saves 30% of administrative time, that 30% should go to teams, not more projects.

4. Train managers to detect weak signals

Technique without human sensitivity is useless. Sensitivity + tools = true protection.

The cost of inaction

If we do not address mental health in AI deployment, we incur three successive bills:

1. Increased absenteeism and turnover, especially among trained young talents.
2. Unachieved AI productivity (exhausted teams do not adopt it).
3. Rising insurance and legal costs (long-term illnesses, daily allowances, psychosocial disputes).

The African and island context

In Mauritius and the region, the subject is less statistically documented but clinically present: management practices inherited from "always more" have yet to account for the cost of mental health. The arrival of AI, without a framework, could accelerate degradation before we measure the impact.

The MOOOVE link

The GMHR program addresses this point precisely: the "Manager Operating Upgrade" explicitly includes a dimension of managerial practices that protect mental health (1:1 rituals, structured feedback, workload management, weak signals), not just technical tools. This is the exact opposite of an AI training focused on "tech first."

Serre is correct: the manager is the primary guardian. The challenge remains to provide them with the means to do so.

Sources: Podcast "Human Footprint" Episode 6, February 17, 2026, with Benoît Serre. News Tank HR, Benoît Serre, on managerial transformation and mental health. Tribune Serre Le Monde, September 2025.

Claudine, for MeetYourJob.