Introducing AI with your works council, not past it
There are two clocks on AI at the workplace, and companies keep watching the wrong one. The EU AI Act’s own duty to inform workers and their representatives before putting a high-risk AI system into use (Article 26(7)) now applies with the rest of the high-risk regime — from 2 December 2027 for most workplace tools after the June 2026 amendments. National codetermination law has no such mercy: in Germany and the Netherlands, works council rights over AI tools apply today, and they can stop a rollout the AI Act would wave through.
Germany: capability is enough
Under the Works Constitution Act, the works council co-determines the introduction of technical systems capable of monitoring employee behavior or performance (§ 87(1) no. 6 BetrVG) — and modern AI tools usually are, whether monitoring was the point or not. Employers must inform the council sufficiently in advance about planned work processes involving AI, selection guidelines created with AI need the council’s consent, and — since the 2021 works council reform — the council may call in an external expert on AI matters, with that consultation deemed necessary. Freshfields’ analysis of employee representatives under the AI Act walks through the interplay.
Netherlands: intent is irrelevant
The Dutch Works Councils Act gives the ondernemingsraad a consent right over systems that process employee personal data or are aimed at or capable of observing presence, behavior, or performance (Article 27(1) WOR) — the employer’s intention doesn’t matter if monitoring could occur as a by-product. A separate advice right covers the introduction of important technological facilities. A practical walkthrough: does the works council need to approve AI tools?
CV screening, AI-assisted appraisals, productivity dashboards, algorithmic scheduling: in both countries, these typically cannot be introduced past the council. Other member states draw the lines differently — the sequencing principle below transfers; the specifics need local advice.
The sequence that works
Companies that struggle treat the council as a sign-off station after vendor selection. The pattern that works inverts it:
- Involve the council before the shortlist. Consent rights bite hardest when positions have already hardened around a purchased tool.
- Share an AI register, not a slide. Which systems, for what purpose, touching which data, with what risk classification. The same inventory your Article 4 evidence rests on is the honest basis for codetermination.
- Agree the guardrails in writing. Purpose limitation, who sees outputs, evaluation dates, and what happens if the tool misbehaves.
- Train the council itself. Council members are asked to approve systems they’ve never operated — and German law’s external-expert right exists precisely because of that gap. AI literacy for employee representatives is a legitimate measure in your program: it speeds every future negotiation, and it is squarely within Article 4’s spirit of literacy for the people dealing with AI decisions.
Why this belongs in your literacy program
A works council that understands AI approves faster and negotiates better guardrails; an employer whose measures visibly include the council’s own literacy signals good faith no slide deck can. When the AI Act’s worker-information duty does arrive in December 2027, the companies that built this muscle will file it under routine.
Our readiness program builds role-based tracks from your policy and tools — employee representatives included — with evidence per person. If your next AI rollout has a council between it and production, book a walkthrough.
Ulern builds readiness and evidence. This post explains the interplay in plain terms — it is not legal advice, and codetermination specifics vary by country and case.