Design and build integrated AI agent solutions in Copilot Studio

Overview

This three-day course provides a comprehensive exploration of designing, building, integrating, and managing AI agent solutions using Microsoft Copilot Studio, with a particular emphasis on the GitHub Copilot harness. Learners will gain practical experience in creating and configuring agents through natural-language instructions, knowledge sources, tools, and reusable skills. The course covers extending agent capabilities via workflows and connections to enterprise systems, as well as enabling collaboration between multiple agents to complete business tasks. Additional topics include testing and evaluating agent behavior, publishing agents, monitoring performance, and understanding the security, governance, and consumption considerations involved in deploying agents within an organization. Hands-on exercises are integrated throughout, allowing learners to configure and connect agents using the low-code Copilot Studio environment.

Prerequisites

Participants should have:

• Practical experience creating, configuring, and testing agents in Microsoft Copilot Studio
• Basic understanding of generative AI and how AI agents use instructions, knowledge sources, and tools
• Familiarity with Microsoft Power Platform concepts, including environments, solutions, and connectors
• Basic familiarity with workflow automation, such as Power Automate or Copilot Studio workflows
• Confidence using Microsoft 365 applications and navigating cloud-based configuration interfaces
• An understanding of authentication, APIs, and structured data is beneficial but not essential
• Programming experience is not required

Target audience

• Power Platform makers interested in expanding their skills into integrated AI agent solutions
• Copilot Studio makers with experience creating agents who wish to explore more advanced capabilities
• Functional consultants and technical business analysts responsible for designing and implementing AI-assisted business processes
• Solution designers and technical specialists aiming to understand multi-agent solutions, agent integration, and operational management

Objectives

By the end of this course, learners will be able to:

• Explain the different agent harnesses available in Copilot Studio and identify suitable approaches for business scenarios
• Design and configure instruction-driven agents using the GitHub Copilot harness
• Extend agents with knowledge sources, tools, and reusable skills to enhance functionality
• Create and integrate workflows that automate business processes and support human decision-making
• Select appropriate methods for integrating agents with enterprise systems, including connectors and MCP servers
• Design multi-agent solutions and configure delegation between connected agents for collaborative tasks
• Test and evaluate agent behavior, including knowledge grounding, tool selection, and delegation
• Publish agents to supported channels and validate their operation within organizational environments
• Monitor agent performance and identify opportunities for improvement
• Explain the security, governance, lifecycle management, and consumption considerations associated with deploying AI agent solutions

Outline

Module 1 – Design and Configure AI Agents

• Understand agent architecture and the GitHub Copilot harness
• Configure agent instructions and behavior to meet business requirements
• Work with environments and solutions in Copilot Studio
• Configure identity, memory, and safety settings for agents
• Create and use reusable agent skills to streamline development

Module 2 – Extend Agents with Workflows

• Understand workflows in Copilot Studio and their role in automation
• Create workflows to automate business processes and tasks
• Work with Dataverse data and Microsoft 365 connectors for integration
• Incorporate approvals and human interaction within workflows
• Test and troubleshoot workflows to ensure reliability

Module 3 – Integrate Knowledge and Tools

• Configure enterprise knowledge sources for agent access
• Connect agents to Dataverse and external services for expanded capabilities
• Explore Model Context Protocol (MCP) servers and their integration
• Configure tools and connector actions to enable agent functionality
• Test knowledge retrieval and tool selection to validate agent performance

Module 4 – Build Multi-Agent Solutions

• Understand multi-agent architectures and their applications
• Create specialist agents for specific business tasks and processes
• Configure agent delegation and collaboration for task completion
• Integrate workflows with specialist agents to support complex scenarios
• Test end-to-end multi-agent scenarios to ensure effective collaboration

Module 5 – Evaluate, Publish, and Manage Agents

• Evaluate agent responses and behavior for quality assurance
• Test knowledge grounding and tool execution to ensure accuracy
• Publish agents to Microsoft Teams and Microsoft 365 channels
• Monitor agent activity and performance for ongoing improvement
• Understand security, governance, and consumption considerations for agent deployment

Exams and assessments

• There is no formal examination or graded assessment included in the course
• Learners consolidate their understanding through practical lab exercises, scenario-based activities, and testing of agent behavior
• Exercises provide opportunities to apply concepts such as agent configuration, workflow automation, enterprise integration, multi-agent collaboration, and agent evaluation

Hands-on learning

• Create and configure Copilot Studio agents using natural-language instructions, tools, and reusable skills
• Build workflows that interact with Dataverse and Microsoft 365 services, including approval processes
• Connect agents to organizational knowledge sources, connectors, and MCP servers
• Create multi-agent solutions by configuring specialist agents and delegation between them
• Test and evaluate agent responses, knowledge retrieval, tool selection, and workflow execution
• Publish agents to supported channels and explore monitoring and management capabilities