You’ve deployed M365 Copilot. Licenses are assigned. Excitement is high, and so is the risk. Once the “toy phase” ends, the real question hits: how do you operationalize AI inside a verified Zero Trust architecture?

This session skips the basics and goes straight to the engineering work. We’ll show how to automate label application using trainable classifiers so Copilot only touches what it should. We’ll break down the Semantic Index and how to influence what it surfaces. And we’ll go deeper into the identity layer, examining Agent 365 and Entra ID’s Agent ID to understand how AI agents authenticate, authorize, and request access in a Zero Trust world.

Finally, we’ll look at how to implement Just‑in‑Time access for AI agents so elevated permissions aren’t left lying around for Copilot (or anyone else) to misuse.

If you want Copilot to behave like a Zero Trust citizen, you need to architect it that way. This session shows you how.

Target Audience: Enterprise Architects & Security Leads who are done with the basics and ready to harden their AI environment.