Build secure and enterprise-ready agents with Agent 365 | BRK251
Neta Haiby explains how to take AI agents to production with security, governance, and observability using the Agent 365 SDK and Microsoft Purview SDK.
Overview
As AI agents move from prototypes to production, this session focuses on the controls teams typically need to make agents enterprise-ready:
- Observability across the agent lifecycle (runtime visibility, activity monitoring, performance metrics)
- Governance (ownership, lifecycle rules, policy-based actions)
- Security and compliance (identity-aware access, data protection)
The session positions Agent 365 SDK as the developer surface for building and instrumenting agents, and Purview SDK as a way to apply governance and compliance controls.
What the session covers
Enterprise requirements for production agents
- Instrumenting agents for runtime visibility
- Enforcing identity-aware access to agent capabilities
- Protecting data used by agents
- Applying policy-based governance without slowing down development
Integration with security ecosystems and third-party agents
- How Agent 365 fits into a broader Microsoft security ecosystem
- Integration patterns for third-party agent platforms
Demo: travel agent built with LangChain (Node.js)
- A practical demo of a travel agent built on LangChain using Node.js
- Using Agent 365 concepts such as skills to help with:
- Identity setup
- Observability instrumentation
Monitoring, ownership, and lifecycle governance
- Unified activity monitoring
- Agent performance metrics
- Managing agent ownership
- Defining rules for automated lifecycle actions
Registry sync and platform integrations
- Demonstration of registry sync with third-party platforms
- Genspark integration with Agent 365, including architecture and security considerations
Speakers and session metadata
- Speakers: Neta Haiby, Kendra Springer, Lei Zhong
- Event: Microsoft Build 2026
- Session: BRK251 | Responsible AI (Breakout, Advanced)