Build secure and enterprise-ready agents with Agent 365 | BRK251

Neta Haiby, Kendra Springer, and Lei Zhong walk through what it takes to ship production-grade AI agents, with practical focus on observability, governance, and security controls that don’t block delivery.

Overview

As AI agents move from prototypes to production, the session focuses on the controls developers and platform teams need across the agent lifecycle:

The presenters show how to build enterprise-ready agents using:

Key concepts covered

Agent lifecycle requirements for production

The session frames “enterprise-ready” agents around five requirements:

Agent Blueprint (reusable instruction model)

A new concept introduced is the Agent Blueprint, described as a reusable instruction model that can be used to create and manage multiple agent instances consistently.

User experience in Microsoft Teams

From the user perspective, the session demonstrates using an agent inside Microsoft Teams, highlighting end-user collaboration scenarios.

Multiple agent instances

The presenters demonstrate how the Blueprint concept supports creating and operating multiple agent instances, and summarize the demo around:

Admin and risk management workflows

The session includes admin-focused capabilities such as:

Rules-based automation

The presenters introduce automation using rules in Agent 365, positioning it as a way to operationalize governance and lifecycle actions.

Resources