Agents are being created faster than most organizations can track them. Some deliver value. Others introduce risk. Some appear outside established processes, with unclear ownership and expanding permissions. This session frames agent governance as an operational response problem rather than a policy discussion. Using Agent 365 as the operational control system, we walk through how organizations can detect, assess, and respond to uncontrolled agent deployment and growth. Over a series of evolving scenarios, the audience becomes part of a simulated response team. As new agent situations emerge, attendees make decisions that affect visibility, access control, lifecycle management, and containment. The outcomes of those decisions are demonstrated in real time. Through practical examples, the session shows how Agent 365 supports monitoring, ownership assignment, access enforcement, and ongoing administration of agents already in production. The focus is on maintaining speed and autonomy for builders while giving organizations the controls needed to operate safely at scale.

Model Context Protocol (MCP) is changing how AI agents interact with tools, share context, and adapt over time. As organizations move beyond static integrations, MCP introduces a new way for agents to reason about available capabilities and coordinate actions. In this session, we demonstrate MCP in practice using Copilot Studio through live demos, starting with straightforward MCP-backed interactions and progressing to more advanced scenarios such as tool discovery, agent and tool chaining, and contextual handoff. Along the way, we show how Copilot Studio communicates with MCP-backed tools, how context is passed and reused, and how these patterns differ from traditional API integrations. Attendees will leave with a practical understanding of where MCP fits in modern agent design and how to apply these patterns in their own environments.

After a challenge has been identified, modern AI solutions start with a choice. Which AI tool? Copilot, Copilot Studio, and Microsoft Foundry each serve different purposes, and the most important decision is where to start and why. As a solution takes shape, that choice is revisited. In this session, we walk through a series of real-world scenarios and pause as requirements evolve to ask new questions: is this still the best fit, should the solution remain focused on a single AI capability, or does it benefit from introducing another? If another capability is added, how do we bring them together in a way that stays effective and manageable rather than creating sprawl? The focus stays on practical decision-making throughout the lifecycle of an AI solution. You will see how thoughtful choices help avoid unnecessary complexity, when adding another AI capability strengthens the outcome, and how to keep solutions maintainable as they grow.

Copilot Studio supports a range of knowledge sources out of the box and works well for many scenarios. As solutions grow, some teams reach a point where they want more control over how information is indexed, searched, and returned to an agent. This is where Azure AI Search becomes a useful extension. In this session, we focus on the why and the how of using Azure AI Search with Copilot Studio. We start by discussing the situations where adding search provides clear benefits, such as large document collections, multiple data sources, or the need for more predictable retrieval. From there, we walk through how the integration works, what decisions matter during setup, and how those choices influence agent responses. Rather than following a single setup recipe, the session highlights practical considerations like index design, data shaping, query behavior, and how Copilot Studio consumes search results. You will see how different choices affect accuracy, performance, and long-term maintainability.