Works Council and AI: Co-Determination by Design

Works councils have co-determination rights when AI systems make decisions about employees. The right approach: implement their requirements as technical constraints in the Decision Layer, not as negotiated compromises in a paper document.

According to the Hans Boeckler Foundation (2024), 68% of works councils in Germany report that AI systems have been introduced in their organizations, but only 31% have a works agreement specifically addressing AI use - leaving the majority of AI deployments without formal co-determination governance.

Works Agreement Requirement	Decision Layer Implementation	Verification
No fully automated performance reviews	Human-in-the-Loop rule (enforced)	Escalation logs in Auditor Portal
Quarterly report on AI-supported decisions	Automated report generation	Portal dashboard with live data
Stop AI on suspected discrimination	Bias monitoring trigger	Alert log and incident history
All controls active and functional	Control_ID per requirement	Evidence history per control

The Usual Approach: Convincing the Works Council

In most companies, AI deployment is treated as a change management project. The works council (Betriebsrat) is informed, convinced, brought along. The goal: getting approval for the works agreement (Betriebsvereinbarung).

This approach has a problem: it treats the works council as a hurdle, not a requirements source. The works agreement ends up as a compromise, written in legal language, filed in a folder. The technical implementation often remains unclear.

The Better Approach: Co-Determination as Architecture Constraint

In the Gosign architecture, works agreements are implemented as technical constraints in the Decision Layer. The works council’s requirements are not negotiated as compromises but implemented as rules in the system.

If the works agreement states: “Decisions on performance reviews must not be made fully automated”, then this is implemented as a Human-in-the-Loop rule in the Decision Layer. The agent physically cannot bypass this rule.

If the works agreement states: “The works council receives a quarterly report on all AI-supported HR decisions”, then the Auditor Portal generates this report automatically.

If the works agreement states: “If discriminatory patterns are suspected, AI usage for the affected process must be stopped immediately”, then this becomes a bias monitoring trigger in the Decision Layer.

What the Works Council Can See

The works council receives access to the Auditor Portal, with a dedicated read-only access restricted to the controls relevant to them.

They can trace: which works agreement rules are implemented as controls, whether all controls are active and functional, how frequently Human-in-the-Loop escalations are triggered, and whether there are anomalies in bias monitoring.

This transparency builds trust, not through promises, but through verifiable technology.

Practical Implementation

Step 1: Analyze existing works agreements. Which provisions affect AI deployment directly or indirectly?

Step 2: Draft a new works agreement for AI deployment. Not as a prose document, but as a requirements catalog with specific, technically implementable rules.

Step 3: Implement requirements as controls in the Decision Layer. Every requirement receives a Control_ID, a technical implementation, and an evidence generator.

Step 4: Grant the works council Auditor Portal access. They can verify at any time whether their requirements are technically implemented.

More on this topic: Co-Determination and AI

Book a meeting. We’ll show you how co-determination is implemented technically.