AI agents become part of your existing IT governance -- not a parallel universe.
AI agents deployed in HR processes, personnel decisions, or employee-relevant workflows are subject to co-determination in Germany. This is not an optional compliance layer -- it is a legal requirement under the Works Constitution Act (Betriebsverfassungsgesetz).
Most AI deployments treat co-determination as an organizational problem: a workshop with the works council, a works agreement, a protocol. That is not enough. When the agent makes decisions affecting employees, co-determination must be technically enforced -- not just agreed upon.
Gosign treats co-determination as an architectural principle. The Decision Layer technically enforces compliance with works agreements. The audit trail documents every decision. The works council can trace what the agent does, why it does it, and when a human intervenes.
The works council has a co-determination right in the "introduction and use of technical facilities designed to monitor the behavior or performance of employees." AI agents that automate HR processes fall under this regulation because they process employee data and make or prepare decisions affecting employees.
The works council must be informed early about the introduction of new technical systems. This applies to the planning phase of AI agents, not just go-live.
Every works agreement is translated into a rule set in the Decision Layer with version ID, validity date, and scope. The agent automatically checks every decision against stored constraints. If a constraint would be violated, the decision is not executed autonomously.
Not a blanket approval process but architecturally controlled routing. Autonomous decision when: high confidence, low risk, no constraint affected. Human-in-the-Loop when: bias risk, co-determination topics, low confidence. The routing cannot be bypassed.
Every agent decision generates a complete audit trail entry: timestamp, agent, input, applied rules (with versions), assessment with confidence and risk score, decision path, output, and status. The works council can view and filter all decisions.
Audit Trail Entry:
├── Timestamp: 2026-02-20T09:14:22Z
├── Agent: hr-merit-cycle-agent
├── Input: Salary adjustment employee #4711
├── Applied Rules:
│ ├── Collective agr.: TV-V, Version 2025.2
│ ├── Works agr.: WA-2024-003, §4 Para. 2
│ └── Salary Band: Band E3, Range 52,000-68,000 EUR
├── Assessment:
│ ├── Confidence: 0.94
│ ├── Risk Score: low
│ └── Result: Within band, within WA constraint
├── Decision Path: autonomous (no WA constraint violated)
├── Output: Booking proposal: adjustment +3.2%
└── Status: approvedAgent Operator: Configures agents, changes rule sets (versioned, documented)
HR Specialist: Views agent decisions within their area of responsibility
Works Council: Insight into decision patterns, constraint status, audit trail (anonymized where necessary)
Auditor / Internal Audit: Full access to audit trail, evidence, control status
Admin: System configuration, no access to professional decision data
RBAC is enforced at database level via Row-Level Security.
Gosign provides templates for: works agreement for AI deployment, technical description for the works council, logging concept, and escalation matrix. These are starting points -- every works agreement must be individually negotiated. Gosign provides the technical foundation, not legal advice.
We do not replace works agreements. We map them technically.
We do not provide legal advice. We deliver the technical infrastructure.
We do not automate decisions subject to co-determination without Human-in-the-Loop.
We do not "train" agents on works council decisions. Agents do not learn uncontrolled.
Co-determination for AI agents is not a workshop topic. It is an architecture topic.
The Gosign architecture technically ensures that works agreements are respected.
The works council can trace what the agent does, why, and when a human intervenes.
Human-in-the-Loop is enforced where required -- architecturally, not optionally.
Agents become part of your existing IT governance -- not a parallel universe.
Yes. AI agents affecting employees are subject to co-determination under §87 BetrVG. The Gosign architecture makes this involvement technically possible, not just organizational.
Every agent decision generates a complete audit trail entry. The Decision Layer documents input, rule, assessment, and result. The works council can trace every decision through the Auditor Portal.
Yes. Works agreements are mapped as explicit constraints in the Decision Layer. Every constraint is versioned and traceable. Changes to works agreements create new constraint versions.
Decision Layer. Human-in-the-Loop. Audit Trail. Works-council-compliant.
Book a Meeting