Overview
Pharmaceutical companies need to assess whether medicines or development targets have sufficient scientific, clinical, competitive, and economic value. Today, such assessments are often conducted manually at selected development milestones, although more frequent and evidence-based insights could improve portfolio, pricing, market, and go/no-go decisions. This project develops a proof of concept (PoC) with a pharmaceutical company in Switzerland for an AI-supported target value assessment system. Using selected publicly known targets or medicines and public data sources, the project will test whether Large Language Models (LLMs) and agent-based workflows can analyze key value dimensions, including biology, feasibility, clinical space, patient benefit, treatment alternatives, competitor activity, market potential, and future-oriented scenarios. The PoC will use specialized AI agents for evidence retrieval, signal extraction, reasoning, synthesis, and quality checking. Outputs will emphasize transparent reasoning, evidence traceability, and expert validation. The project will deliver a PoC, benchmark and feasibility report, and a solid basis for future research.
