AI and social dialogue: from defensive negotiation to a discussion about means

In most companies, a works-council consultation on an AI project follows the same script: leadership presents reassuring, general intentions, employee representatives respond with the fears that sector studies have been feeding for months, and both sides end up negotiating on ground neither can actually verify. This isn’t a goodwill problem. It’s a data problem. Here’s how a factual, role-by-role baseline changes the nature of that discussion.

The blind spot in AI-related social dialogue

A works-council presentation on an AI project usually looks more like a communications exercise than a data briefing: “AI will help us stay competitive,” “no forced layoffs are planned at this stage,” “we’ll support the teams affected.” None of these statements is false, but none of them says anything verifiable — how many roles, in which division, on what timeline, with what reskilling budget. On the other side, employee representatives often have nothing to argue with beyond the same macro sector statistics leadership carefully avoids citing, because they apply no more precisely to the company than the talking points presented in the room.

The result: a negotiation where neither side holds a number it can defend, and where trust gets built — or destroyed — on the tone of the message rather than its substance.

What the law already requires, and what’s coming

The regulatory framework leaves little room for vagueness anymore. In France, the triennial GEPP workforce-planning negotiation requires, for any company above 300 employees, a works-council consultation whenever a transformation touches the organization of work — which covers, in practice, most AI projects with an impact on roles. On top of that, the EU AI Act imposes documentation obligations on high-risk uses starting August 2026, and December 2027 for entities covered by the Omnibus mechanism. And that’s before accounting for data-protection requirements whenever a project processes data about employees themselves.

These aren’t peripheral compliance boxes to tick. They create a genuine need for traceability and justification that a general communication plan simply cannot satisfy — it takes precise, documented, and regularly updated data.

A social-dialogue negotiation without precise data isn’t fought on facts. It’s fought over who tells the more convincing story — management or employee representatives — and that serves neither side well.

What a factual baseline changes in the negotiation

• Scope becomes precise — instead of “some roles could be affected,” leadership can state which tasks, in which division, represent what share of automatable time, with a verifiable order of magnitude.
• Timeline becomes something to discuss — a realistic schedule, grounded in a diagnostic, replaces vague phrasing like “in the medium term,” which commits nobody and reassures even less.
• The reskilling budget gets calculated, not announced — knowing the nature of the tasks shifting lets you size real training needs, instead of announcing an arbitrary envelope.
• Follow-up becomes a verifiable commitment — a quarterly measurement point lets employee representatives see actual change over time, rather than having to take a one-off promise on faith.

Three steps for approaching the works council with data instead of intentions

1. Diagnose before announcing — measure real, role-by-role exposure before any public communication. Presenting a project before you have the numbers means improvising a negotiation you could have prepared instead.
2. Present by role, not by an aggregate percentage — a single overall figure always hides very different situations from one division to the next. Even partial detail inspires more trust than an average that papers over the gaps.
3. Announce the next measurement point — stating upfront that the diagnostic will be revisited in the coming months shows the topic stays steered over time, rather than being treated as closed after the first presentation.

Vague over-communication is a risk too

Many leadership teams think staying deliberately vague about the scope of a project limits tension, out of fear of alarming people. That’s a risky bet: a vague message that turns out, six months later, to be at odds with what people observe on the ground costs more in trust than a precise one — even one that announces a bigger change than expected. Employee representatives who catch a gap between the message and the facts start negotiating everything afterward, including unrelated topics, with a distrust that’s far harder to dissolve.

What a leadership team gains by choosing costed transparency

A leadership team that shows up with a factual diagnostic, built from the company’s own data — without that data ever passing through third-party systems, under a sovereign processing approach that meets data-protection requirements — changes the shape of the negotiation itself. The discussion stops being about whether the project is real, which is contestable without evidence, and starts being about the means of support, which are genuinely negotiable. That’s a shift that benefits both sides: leadership gains credibility, and employee representatives gain a discussion partner working from stable ground instead of one hiding behind talking points.

Social dialogue on AI will never be a smooth ride. But it can stop being a negotiation in the dark — provided the data arrives before the talking points, and not the other way around.