The use of AI programmes in companies and administrations requires an active labour policy that gives employees and their representatives the opportunity to be involved throughout the entire introduction cycle of artificial intelligence (AI), as is also regulated by Section 90 (1) No. 3 BetrVG (right to information). This support has been given a further legal basis by the EU-AI Act of August 2024. The introduction of artificial intelligence (AI) requires the participation and qualification of employee representatives and employees who are likely to be confronted with AI applications in the workplace. What should the AI model be able to do, who bears responsibility and liability? Does the artificial intelligence (AI) support the employees or does the artificial intelligence (AI) act independently?

The first step (1), taking into account Section 90.2 BetrVG (right to consultation), involves a joint understanding of the objectives that the company associates with the AI model, followed by a situation assessment, the joint determination of analytical assessment criteria and the selection of a pilot area in the company (Section 91 BetrVG). In the event of changes to the job profiles in the pilot area, if the knowledge and skills of the employees are no longer sufficient to fulfil the task, co-determination applies in accordance with Section 97 (2) BetrVG.

The associated vocational training measures must be approved by the works council (BR) in accordance with Section 98 (1) BetrVG. The second step (2) involves a joint understanding of data. What are the data sources, what is the quality of the data, is the data complete and what information, classes and cases can be obtained from the data. This aspect also concerns ethical aspects and touches on Section 87 (1) No. 6 BetrVG, i.e. this process is subject to co-determination. The following step (3) is the most time-consuming and concerns the process of data preparation, the selection of data, the transformation into usable data formats, the artificial expansion of the data and finally the decision as to which data should be used for training, validation and testing of the artificial intelligence (AI).

Once this process is largely complete, the next step (4) involves selecting the artificial intelligence (AI) models to be used in the company, whereby the data is now tested on various models. The fifth step involves an initial evaluation of whether the selected model has achieved the objective of the AI implementation. If not, why? If the AI application was not successful and the objectives were not achieved, management and the works council agree to return to the first process step. However, if the objectives in the pilot area have been successfully achieved, the final step (6) involves implementing the AI application in day-to-day operations with continuous process review. This AI design process is very training-intensive for both the employees and the interest groups; it has been mandatory for companies since February 2025 due to the regulations of the EU AI ACT. This multi-stage participation concept was developed by the center of cooperation RUB/IGM as part of the regional competence centre ‘humaine’ at Ruhr-Universität Bochum (RUB), which is funded by the Federal Ministry of Education and Research (BMBF).