GEPP and AI: consulting the works council with data, not estimates

A GEPP negotiation that approaches AI without precise data starts off on the wrong foot. Leadership puts forward general intentions about “the impact of artificial intelligence on jobs,” employee representatives receive them with justified suspicion, and the discussion bogs down in impressions nobody can verify. Above 300 employees, this negotiation isn’t optional: the triennial GEPP agreement requires a works-council (CSE) consultation whenever a transformation touches the organization of work, and AI unquestionably qualifies. Here’s how a role-by-role diagnostic changes the nature of that consultation — and why going without one gets expensive.

What GEPP actually requires above 300 employees

France’s triennial workforce and skills planning framework (GEPP) governs, every three years, how a company anticipates the evolution of its roles and skills. It isn’t a formality: as soon as the organization of work is affected — and introducing AI tools that change tasks, volumes or role boundaries almost always qualifies — French labor law requires an information-consultation process with the works council. A company that treats AI as a purely technological question, without anticipating this social dimension, often discovers the obligation only once it’s already in breach of it.

The classic blind spot: negotiating on impressions

The scenario repeats itself across companies: HR shows up to the negotiation with cautious phrasing — “certain roles could be impacted,” “we’re studying possible use cases” — for lack of an actual map. Employee representatives, reasonably, read that vagueness as withheld information rather than genuine uncertainty. Trust erodes before the first concrete proposal ever hits the table.

That vagueness has a direct cost: it opens the door for the works council to mandate an external accounting expert to assess the project’s impact itself, funded by the employer, on a timeline that can add months to the negotiation. A leadership team that doesn’t measure its own role exposure leaves someone else to do it instead, with oversight it no longer controls.

What a role-by-role diagnostic brings to the table

A diagnostic that breaks each role down into tasks, checks each one against a reference library, then consolidates results division by division, changes the nature of the exchange. Leadership no longer shows up with intentions but with a dated snapshot: how many roles are affected, to what degree, on what horizon, with what acknowledged margin of uncertainty. That snapshot doesn’t remove the need to negotiate — it changes what’s being negotiated, from “is this true?” to “what do we do about it?”

A GEPP negotiation that opens without company-specific data isn’t really about AI. It’s about how much trust the room is willing to extend to leadership’s word alone.

Three uses of the diagnostic across the GEPP cycle

The diagnostic isn’t used the same way at every stage of the three-year cycle:

1. Ahead of the negotiation — an org-wide view, division by division, lets HR open the discussion with a factual baseline instead of an intuition to defend.
2. During the negotiation, division by division — a deeper look at the most affected divisions allows precise answers to works-council questions about a specific department, instead of falling back on generalities.
3. After the agreement, every quarter — a recurring cycle documents, throughout the three years of the agreement, that the commitments made are tracked with the same data used to negotiate them — not data that went stale the day it was signed.

The cost of showing up empty-handed: works-council expertise and lasting distrust

A works council with no reliable data from management has essentially one reasonable lever left: request an external expert review. That review, funded by the employer for any major project changing working conditions, delays the negotiation by months and puts leadership on the defensive for the rest of the cycle. The financial cost is real; the trust cost is worse, because it plants the idea that leadership is discovering AI’s impact at the same time as everyone else.

What the data protects on the employee side too

The value of a quantified diagnostic isn’t one-sided. A precise, role-by-role map lets employee representatives distinguish divisions genuinely exposed in the short term from those that aren’t — instead of negotiating blanket protective measures that spread effort thin across roles that need it least, at the expense of those that need it most. A training or internal-mobility plan built on precise data protects the people actually affected better than a generic plan built on fear.

For a leadership team opening a GEPP negotiation in 2026 or 2027, the question is no longer whether AI belongs on the agenda. It’s whether that negotiation runs on what the company knows — or on what it merely believes.