Bayer built PRINCE, an AI research assistant that transforms how scientists search preclinical drug data—shifting from keyword search to natural language questions, then to autonomous task execution.

What happened: Bayer developed PRINCE (Preclinical Information Center), an agentic AI system built on Retrieval-Augmented Generation (RAG—a technique that combines AI text generation with precise information retrieval). The platform evolved through three phases: Search (unified gateway to structured study metadata), Ask (natural language question-answering over unstructured PDF reports), and Do (multi-agent system that executes complex research tasks). The system uses LangGraph for orchestration and FastAPI for serving, with a React UI for researchers to submit queries.

Why it matters: Preclinical researchers at large pharmaceutical organizations face fragmented data scattered across many systems, inadequate keyword-based search tools, and time-consuming manual document review. PRINCE addresses these by enabling researchers to pose complex questions in plain language and receive answers grounded in decades of accumulated study reports—potentially freeing up researcher time and enabling faster, data-driven decision-making in drug development.

What to watch: The system includes deliberate pause points, feedback loops, and validation steps to ensure data completeness before responding. PRINCE now operates as an active research assistant capable of drafting regulatory documents and orchestrating workflows, representing a shift from passive information retrieval to active task execution in preclinical development.