Dependable Integration Concepts for Human-Centric AI-Based Systems