AI Operations to Own the Fleet Master the Mission in Microsoft Foundry - BRK205

Learn to master AI operations at enterprise scale through this comprehensive conference session from Microsoft Ignite 2025. Discover Microsoft's two-dimensional security approach for AI agents and explore how Foundry Control Plane provides real-time visibility, governance, and optimization capabilities for AI fleet management. Understand the critical risks associated with AI agents accessing both external and internal systems, and learn to implement effective guardrails with controls for prompt injection and PII protection. Explore observability frameworks and human oversight mechanisms essential for trustworthy AI deployment, including continuous evaluations on real production data. Examine security integration with Microsoft Defender for comprehensive threat detection and learn to manage security and policy alerts within Foundry's observability platform. Gain insights from a real-world customer case study featuring Stanley Black and Decker's implementation. This intermediate-level breakout session is presented by Ken Archer, Sarah Bird, and Florent Ricci, providing practical strategies for monitoring, optimizing, and governing AI agents securely while ensuring reliability and compliance in production environments.

Syllabus

0:00 - Microsoft's Two-Dimensional Security Approach for AI Agents
00:07:02 - Introduction to Foundry Control Plane's Capabilities
00:09:06 - Understanding Agent Risks: External and Internal World Access
00:13:47 - Creating a guardrail and adding controls for prompt injection and PII protection
00:17:28 - Introduction to observability and human oversight in trustworthy AI
00:23:03 - Introduction to Continuous Evaluations on Real Production Data
00:28:35 - Security Integration with Microsoft Defender for Threat Detection
00:34:45 - Discussing security and policy alerts in Foundry observability
00:36:49 - Introduction of customer case study: Stanley Black and Decker