In this Microsoft Ignite 2025 session, Derah Onuorah and Wassim Wehbi share practical insights on building advanced AI agents using Copilot Studio. The video guides intermediate users through agent types, integration with Model Context Protocol, and demos of human-in-the-loop workflows.

Deep Dive into AI Tools in Copilot Studio

Session Speakers: Derah Onuorah, Wassim Wehbi
Source: Microsoft Ignite 2025
Session Code: BRK315

Overview

This session explores how to use the latest industry- and function-led AI tools available in Copilot Studio to create powerful agents for Microsoft 365 Copilot and other Microsoft cloud environments. It provides an in-depth look at agent capabilities, human-in-the-loop features, workflow integration, and practical use cases for both technical and business solutions.

Chapters & Key Topics

  • What is Copilot Studio – Three Experience Types (0:00)
    • Introduction to the key experiences supported in Copilot Studio
  • Types of Agents – Retrieval, Task, and Autonomous Agents (3:34)
    • Different agent archetypes: retrieval-based, workflow-driven, and fully autonomous
  • Using Model Context Protocol (MCP) with Copilot Studio (8:22)
    • Deep dive on leveraging the MCP framework to improve agent context and intelligence
  • Tools and Connectors for Agent Integration (13:08)
    • Overview of available connectors and integrated functions within the agent toolchain
  • Human-in-the-Loop Request for Information (RFI) in Agents (20:11)
    • Demonstrating approaches for embedding humans for critical review or intervention
  • Agents vs. Workflows: When to Use Each (24:52)
    • Decision-making guide for selecting agents or classic workflows for automation tasks
  • AI Prompts for Extracting Invoice Information (29:25)
    • How advanced prompting can streamline document AI scenarios, like invoice extraction
  • Agent Demos (Full Scenario With Human-in-the-Loop and AI Decision Making) (32:33–38:29)
    • Complete demonstration of an agent in action, including RFI and automation hand-off

Practical Takeaways

  • Agent Types: Learn which agent forms—retrieval, task, autonomous—fit which business scenarios.
  • MCP Use: Enhance your agents by providing them with richer contextual data using Model Context Protocol.
  • Low-Code Paths: Build workflow automation and AI-powered bots without deep coding.
  • Integrations: Connect agents easily to enterprise data, APIs, and Microsoft 365 services.
  • Human-in-the-Loop: Build solutions that check with a human when appropriate, balancing automation and oversight.

Resources

Tags

Copilot Studio, AI agents, Microsoft 365, Model Context Protocol, Low-code development, Human-in-the-loop, RFI, Workflow automation, Autonomous agents, Connectors, Microsoft Ignite, Business solutions, Agent demo, AI decision making, Retrieval agents, Task agents

About the Speakers

Derah Onuorah and Wassim Wehbi are experienced product leaders in the AI and Microsoft low-code solution space, presenting as part of Microsoft Ignite’s 2025 expert session lineup.