Microsoft Developer brings together Uli, Eric, and David—Armchair Architects—to explore multi-agent orchestration in enterprise environments, covering patterns, frameworks, and security practices for building secure and scalable AI agent solutions.

Armchair Architects: Patterns and Best Practices for Multi-Agent AI Orchestration

In this episode of the Azure Essentials Show, Uli Homann, Eric Charran, and David Blank-Edelman—known as the Armchair Architects—dive deep into the challenges and opportunities of building multi-agent architectures in the enterprise. Their conversation covers:

Key Topics Covered

  • Architecting Multi-Agent Systems:
    • Approaches for orchestrating multiple AI agents, including patterns such as specialist delegation, sequential workflows, parallel reasoning, message-based inter-agent communication, and consensus building.
    • Examples of agent patterns (e.g., reviewer, negotiation, and employees-as-agents).
  • Tools and Frameworks:
  • Security and Data Governance:
    • Critical need for data leak prevention (DLP), policy enforcement, and governance practices—especially when agents communicate autonomously or access sensitive enterprise data.
    • How Microsoft Purview helps with compliance and managing sensitive data.
  • Cost Management and Practical Concerns:
    • Considerations for model selection, monitoring operational costs, and ensuring scalability.

Actionable Guidance

  • Review and implement message-based and delegation architectural patterns as a strong foundation for multi-agent orchestration.
  • Experiment with open-source frameworks (AutoGen, Semantic Kernel) to understand multi-agent workflows in practice.
  • Assess and enforce data governance and security controls—including DLP mechanisms—whenever deploying agent systems in production environments.

Learn More

Hosts