Agentic Science is an emerging paradigm where autonomous Artificial Intelligence (AI) agents communicate directly with structured biomedical data and services to support discovery, diagnostics, and care. This session will explore how GA4GH standards — such as Data Connect, Passport, WES, and TES — can enable AI agents (e.g. LLM-powered copilots or task-based bots) to reason over distributed -omics and health datasets. We’ll showcase early prototypes where agents generate, execute, and refine standard-compliant queries across federated networks, and discuss challenges such as schema alignment, explainability, and security. The goal is to catalyse cross-Work-Stream collaboration around an agentic reference architecture and roadmap for enabling AI to act as an intelligent interface between people and data.