In the "SATW Insights" column on inside-it, Manuel Kugler, Programme Manager Data & AI and Advanced Manufacturing at SATW, makes it clear that anyone looking for guidance on the responsible use of AI today will find plenty of material - from the EU AI Act to companies' own Ethical AI Principles. What is missing is the translation into concrete terms. For example, what does "transparency" mean for an IT specialist who is to introduce a new AI system?
Using four company examples, the article provides answers and distils five key recommendations:
- Actively involve stakeholders - not just inform them, but integrate them into the development process. Those who can express concerns contribute to success.
- Ethical evaluation right from the start - don't wait until the go-live, but ask critical questions as early as the data selection and model design stages.
- From principles to measurable criteria - instead of vague demands such as "The system should be fair", verifiable requirements are needed, such as: "The false positive rate must not deviate by more than five per cent between groups."
- Implement transparency in technical terms - make explainability a design decision rather than lip service.
- Understanding responsibility as an ongoing process - responsible use of AI requires more than good intentions; it must be systematically embedded in structures and processes.
Read article in German
