Session (CTI)

Issue: Retrieval-Augmented Generation (RAG) is rapidly becoming a cornerstone for grounding evidence in HEOR workflows. By anchoring outputs in sourced literature and real-world data, RAG ensures that evidence generation is both transparent and traceable. Multi-agent architectures build on this foundation by orchestrating specialized agents to collaboratively process complex HEOR tasks, in which each agent has distinct roles such as document retrieval, synthesis, validation, and communication. This modular approach allows for consistency and reliability across the entire evidence generation pipeline, as each agent can leverage RAG to verify information and maintain a clear chain of reasoning. As HEOR processes scale, interoperability emerges as a critical enabler. Seamless communication among diverse agents and platforms ensures that data, context, and reasoning can be exchanged without loss of fidelity or meaning. The Model Context Protocol (MCP) addresses this challenge by offering a standardized framework for sharing contextual information, evidence, and logical reasoning among systems. MCP supports reproducibility, facilitates integration with regulatory workflows. This panel will present a comprehensive overview of how multi-agent architectures, RAG, and MCP have the potential to reshape the landscape of evidence generation in HEOR. System designers will share practical insights on interoperable, modular AI-driven solutions, while an HTA representative will discuss expectations around transparency, governance, validation, and responsible AI use. Attendees will gain an understanding of the opportunities and governance perspectives for deploying these advanced architectures. Overview: (5 min) Bill Malcolm: Introduction to multi-agent AI, interoperability, and MCP in HEOR/HTA. (10 min) Rajdeep Kaur: Agentic AI and RAG: how agentic systems interact with RAG for evidence retrieval, grounding, and workflow integration. (15 min) Sven L Klijn: Multi-agent architectures: designs, coordination, and interoperability challenges. (15 min) Daniel Ollendorf : HTA perspective: expectations for interoperable, agentic systems. (15 min) Audience Q&A