Building Reliable Agentic AI Systems

Bayer AG and Thoughtworks developed PRINCE, a cloud-based agentic AI system that transforms pharmaceutical preclinical research by replacing traditional keyword searches with an intelligent assistant capable of answering complex questions across decades of safety study reports. The platform uses Retrieval-Augmented Generation and Text-to-SQL with specialized agents for clarifying intent, planning, research, validation, and document synthesis, while prioritizing reliability through transparency, explainability, and human oversight. PRINCE demonstrates how context engineering and harness engineering—managing information routing and system orchestration—enable production-ready AI systems that improve data accessibility and research efficiency while maintaining governance and compliance in regulated industries.