Session Abstract

What A demo-driven session showing how to design and orchestrate multi-agent systems using the Microsoft Agent Framework — moving beyond single-agent solutions into coordinated agent networks that handle complex enterprise workflows reliably in production. Why Single agents have limits. Complex workflows spanning multiple systems and requiring specialised reasoning demand a different architectural approach. Most organisations attempting multi-agent architectures do it without a structured framework — resulting in brittle systems that work in demo but fail in production when agents loop, disagree, or produce outputs that cannot be audited. This session shows the right way — with honest coverage of failure modes and governance patterns most conference sessions never address. How - Architecture overview — orchestrator and specialist agents, communication protocols, handoff mechanisms, shared context management, and failure handling - Live demo of multiple agents orchestrated against a real enterprise workflow — delegation, parallel execution, context passing, and live tool invocation demonstrated end to end - Honest coverage of production failure modes — infinite loops, agent disagreement, timeout scenarios, and graceful degradation - Q&A on architecture decisions, governance, cost management, and real-world implementation challenges Who Solution architects, enterprise developers, and technical leads designing multi-agent AI systems. Solid experience with Copilot Studio, Azure AI Foundry, or the Microsoft 365 Agents SDK required. Skill level: Advanced. Top 3 Key Takeaways 1. A complete architectural understanding of multi-agent orchestration — communication patterns, handoff mechanisms, and failure handling that make multi-agent systems reliable in production. 2. First-hand experience seeing multiple specialised agents working together live on a real enterprise workflow including delegation, parallel execution, and context passing. 3. A practical governance and production readiness framework covering monitoring, auditability, and the failure points that only reveal themselves when real users interact with a live multi-agent system.

Samir Makwana

Samir Makwana empowers enterprises and government agencies to harness the full power of Microsoft Cloud—and especially its cutting-edge AI stack—to build secure, value-driven businesses. With 17 years as a software developer, pre-sales engineer, and solution architect, he’s architected end-to-end IT transformations that blend security, automation, and collaboration. A true AI evangelist, Samir designs PoCs and production systems leveraging Azure AI services (Cognitive Services, Azure OpenAI, Semantic Kernel), Copilot Studio, and Power Platform AI—applying deep prompt engineering, vector search/RAG patterns, and grounding techniques to deliver next-gen user experiences. He’s integrated generative AI agents into Office 365 and custom apps, automated workflows with Power Automate and Logic Apps, and secured data with Zero Trust and Microsoft Purview. As Founder & Principal Architect of Cloudtrify Inc., Samir crafts both short- and long-term cloud-native and AI solutions that drive ROI, user adoption, and maintainability—always with responsible AI and compliance top-of-mind.