Clinical AI Orchestration & Evidence-Based Practice — RGMBS 2026
Session 02 at the Riyadh Global Medical Biotechnology Summit 2026, presented by Dr. Fady Hannah-Shmouni, MD FRCPC — Endocrinologist, Professor at the UBC Department of Medicine, Director of AI at GCLS, and Founder & CEO of Hormonaly.ai. A hands-on architect workshop, roughly 2h 45m, pods of 4–6, no coding required for the core track.
The shift to agentic clinical intelligence
The move from passive AI assistance to agentic clinical intelligence is the next multiplier in life sciences. Participants act as architects, building the connective tissue between frontier models and real-world clinical practice.
The clinical AI stack — four layers
- Model layer — Claude Scientific for complex research queries, alongside Saudi Arabia's ALLaM for localized, Arabic-native clinical intelligence, plus open and sovereign options self-hostable in-Kingdom.
- Orchestration & RAG layer — multi-agent coordination with LangGraph, a "Second Brain" knowledge-management system, and retrieval-augmented generation for clinical grounding. Tool interoperability over MCP.
- Clinical tools layer — OpenEvidence for real-time evidence synthesis, Heidi Health for ambient documentation, and Hormonaly for metabolic and endocrine clinical modelling, wired directly into the agent loop.
- Domain & governance layer — guideline grounding, on-source enforcement, verification loops, provenance, audit trails and a human in the loop.
What you take home
A functional clinical AI agent that orchestrates evidence synthesis, patient documentation and expert clinical reasoning; the manifest that defines it; the governance pattern; and an MCP connector to plug it into your own tools.