Slides presented at the EFMC2 Meeting, 4 September 2026
https://www.drugdiscovery.net/data/20260904_Agents_v4.pptx
Copilot summary: ‘The presentation argues that the true value of AI agents in drug discovery lies in improving clinically relevant decisions rather than simply increasing automation, data volume, or workflow efficiency. Agents are described as combinations of models and tool-calling systems that can effectively support information gathering, analysis, and experimental workflows, but their greatest challenge remains high-precision decision making in complex biological spaces. Through examples including toxicity prediction, drug repurposing, virtual scientists, and laboratory automation, the speaker highlights that predictive validity, domain knowledge, and carefully chosen heuristics are more important than scale alone. The central message is that successful drug discovery requires integrating data, scientific understanding, and useful biases to guide decisions, since data and AI by themselves are insufficient to navigate the vast and uncertain search spaces of biology and medicine.’