This course develops the knowledge and skills required to design, build, integrate, secure, and deploy Microsoft Copilot agents for enterprise use. It covers Microsoft 365 Copilot Agent Builder and Microsoft Copilot Studio, progressing from agent instructions, conversation design, and knowledge grounding to tools, agent flows, REST APIs, generative orchestration, autonomous agents, Model Context Protocol (MCP), and multi-agent architecture.

Participants explore human approval controls, security, governance, monitoring, and application lifecycle management. The course concludes with a capstone that brings these capabilities together in an enterprise agent solution.

 

Duration 5 Days – 35 hrs.

 

Objectives

• Distinguish generative AI, agentic AI, conversational agents, autonomous agents, and multi-agent systems.
• Select an appropriate Microsoft Copilot development approach for a business requirement.
• Create and configure agents using Microsoft 365 Copilot Agent Builder and Microsoft Copilot Studio.
• Write clear agent instructions, task boundaries, response rules, and escalation guidance.
• Design reusable topics, conversation flows, variables, and topic inputs and outputs.
• Configure knowledge sources and produce grounded responses with source attribution.
• Integrate connectors, agent flows, REST APIs, and MCP tools.
• Configure generative orchestration for tool selection and multi-step task execution.
• Design autonomous and multi-agent solutions with appropriate approval controls.
• Apply authentication, authorization, least privilege, data protection, and governance practices.
• Test agent behavior, investigate failures, and monitor solution performance.
• Publish and manage agents across development, test, and production environments.

 

Target Audience 

• Power Platform developers and automation engineers.
• Microsoft 365 developers and application developers.
• Solution architects and technical leads.
• Integration developers working with enterprise APIs and business systems.
• Technical consultants designing Microsoft Copilot solutions.
• Administrators and technical specialists responsible for agent deployment and governance.

 

Prerequisites 

• Familiarity with Microsoft 365, including Teams and SharePoint.
• Basic experience with Microsoft Copilot Studio or comparable conversational agent development.
• Working knowledge of Power Platform and Power Automate.
• Understanding of generative AI concepts and basic prompt design.
• Familiarity with REST APIs, HTTP methods, and JSON.
• Basic understanding of authentication, authorization, and cloud environments.
• Access to a suitably licensed Microsoft environment with the permissions required to create agents, configure connections, and publish solutions.

 

 

 

Course Outline 

Day 1 — Agentic AI and Microsoft Copilot Agent Fundamentals

Module 1 — Understanding Agentic AI 

• Introduction to Generative AI and Agentic AI
• Generative AI vs. Agentic AI
• Chatbots vs. Copilots vs. AI Agents
• Characteristics of an AI Agent
• Goals and Instructions
• Reasoning and Planning
• Knowledge and Grounding
• Tools and Actions
• Memory and Context
• Observation and Feedback
• Human-in-the-Loop
• Conversational vs. Autonomous Agents
• Single-Agent vs. Multi-Agent Systems

 Module 2 — Microsoft Copilot Agent Ecosystem 

• Microsoft 365 Copilot
• Microsoft 365 Copilot Agent Builder
• Microsoft Copilot Studio
• Microsoft Power Platform
• Power Automate and Agent Flows
• Microsoft Dataverse
• Microsoft SharePoint
• Microsoft Teams
• Microsoft Graph Concepts
• Microsoft Connectors
• REST APIs
• Model Context Protocol (MCP)
• Microsoft Azure Services
• Choosing the Appropriate Copilot Development Approach

 Module 3 — Building Microsoft 365 Copilot Agents

• Creating Agents with Microsoft 365 Copilot
• Agent Name and Description
• Defining Agent Instructions
• Configuring Knowledge Sources
• Creating Starter Prompts
• Testing Agents
• Sharing Agents
• Understanding Agent Builder Capabilities and Limitations
• Moving from Agent Builder to Copilot Studio

Day 2 — Building Agents with Microsoft Copilot Studio

Module 4 — Microsoft Copilot Studio Fundamentals 

• Introduction to Microsoft Copilot Studio
• Copilot Studio Environment
• Creating an Agent
• Agent Configuration and Overview
• Agent Instructions
• Knowledge
• Topics
• Tools
• Variables
• Entities
• System Topics
• Testing Panel
• Activity Map
• Publishing Agents

 Module 5 — Designing Effective Agent Instructions 

• Understanding Agent Instructions
• Role Definition
• Goals and Task Boundaries
• Tool Selection Guidance
• Knowledge Selection Guidance
• Response Rules
• Agent Constraints
• Handling Ambiguous Requests
• Escalation Rules
• Instruction Conflicts
• Prompt and Instruction Design Best Practices
• Common Instruction Design Mistakes

 Module 6 — Topics and Conversation Design 

• Understanding Topics
• System Topics vs. Custom Topics
• Trigger Phrases
• Trigger Nodes
• Message Nodes
• Question Nodes
• Condition Nodes
• Tool Nodes
• Redirect Nodes
• Variables
• Topic Inputs and Outputs
• Reusable Topics
• Conversation Flow Design
• Classic vs. Generative Orchestration

 Module 7 — Knowledge and Generative Answers 

• Understanding Grounding
• Generative Answers
• Knowledge Sources
• SharePoint Knowledge
• Website Knowledge Sources
• Uploaded Documents
• Microsoft Dataverse
• Enterprise Data
• Knowledge Descriptions
• Knowledge Selection
• Source Attribution
• Grounded vs. Ungrounded Responses
• Reducing Hallucinations
• Knowledge Security Considerations

Day 3 — Tools, Actions, APIs and Agent Flows

Module 8 — Tools and Actions 

• Understanding Agent Tools
• Tools vs. Knowledge
• Tool Inputs and Outputs
• Tool Descriptions
• Tool Selection
• Microsoft Connectors
• Custom Connectors
• REST APIs as Agent Tools
• Tool Authentication
• Tool Error Handling
• Tool Permissions
• Designing Safe Agent Actions

 Module 9 — Agent Flows 

• Introduction to Agent Flows
• Agent Flow Architecture
• Flow Inputs and Outputs
• Conditions
• Variables
• Connectors
• Calling Flows from Agents
• Returning Results to Agents
• Error Handling
• Reusable Agent Flows
• Deterministic Workflows vs. AI Decisions
• Integrating Agent Flows with Copilot Studio

 Module 10 — Connecting REST APIs 

• REST API Fundamentals
• HTTP Methods: GET, POST, PUT and DELETE
• JSON Requests and Responses
• API Authentication
• API Keys
• OAuth Concepts
• Custom Connectors
• Request and Response Mapping
• OpenAPI Concepts
• API Error Handling
• Integrating Enterprise APIs with Copilot Agents

Day 4 — Generative Orchestration, Autonomous Agents and MCP

Module 11 — Generative Orchestration 

• Introduction to Generative Orchestration
• Generative vs. Classic Orchestration
• Understanding Agent Planning
• Dynamic Tool Selection
• Dynamic Knowledge Selection
• Topic Selection
• Chaining Actions
• Multi-Step Execution
• Agent Descriptions
• Tool Descriptions
• Inputs and Outputs
• Replanning
• Agent Guardrails
• Testing Generative Orchestration

 Module 12 — Autonomous Agents 

• Introduction to Autonomous Agents
• Conversational vs. Autonomous Agents
• Event-Driven Agents
• Autonomous Triggers
• Background Processes
• Instructions and Decision Boundaries
• Autonomous Actions
• Agent Guardrails
• Approval Requirements
• Auditability
• Designing Safe Autonomous Agents

 Module 13 — Model Context Protocol (MCP)

•  Introduction to Model Context Protocol
• Why MCP Is Used
• MCP Architecture
• MCP Clients
• MCP Servers
• MCP Tools
• MCP Resources
• MCP Prompts
• Tool Discovery
• Connecting Enterprise Systems
• MCP vs. Microsoft Connectors
• MCP vs. REST APIs
• MCP Security Considerations
• Using MCP with Copilot Agents

 

Day 5 — Multi-Agent Systems, Security, Governance and Production

Module 14 — Multi-Agent Architecture 

• Introduction to Multi-Agent Systems
• Single-Agent vs. Multi-Agent Architecture
• Agent Specialization
• Parent and Child Agents
• Agent Delegation
• Agent Handoffs
• Agent Descriptions
• Agent Routing
• Shared Business Processes
• Multi-Agent Design Patterns
• Agent Coordination
• When to Use Multi-Agent Architecture
• When Not to Use Multi-Agent Architecture

 Module 15 — Human-in-the-Loop 

• Human-in-the-Loop Concepts
• Approval Gates
• Risk Classification
• Confidence and Decision Boundaries
• Escalation
• Sensitive Operations
• Human Review
• Business Rules
• Agent Autonomy Boundaries
• Balancing Automation and Human Control

 Module 16 — Security and Governance

• Authentication
• Authorization
• User Identity
• Connection Permissions
• Principle of Least Privilege
• Data Loss Prevention (DLP)
• Environment Strategy
• Connector Governance
• Prompt Injection
• Indirect Prompt Injection
• Data Leakage
• Tool Abuse
• Excessive Agency
• Protecting Sensitive Actions
• Auditability
• Agent Governance
• Enterprise Security Best Practices

 Module 17 — Testing, Evaluation and Monitoring

• Testing Copilot Agents
• Test Cases
• Test Sets
• Agent Evaluation
• Response Evaluation
• Tool Selection Testing
• Knowledge Accuracy Testing
• Topic Testing
• Orchestration Testing
• Hallucination Testing
• Edge-Case Testing
• Agent Analytics
• Monitoring
• Performance Improvement
• Continuous Agent Improvement

 Module 18 — Publishing and Enterprise Deployment

• Publishing Copilot Agents
• Deploying to Microsoft Teams
• Deploying to Microsoft 365 Copilot
• Deploying to SharePoint
• Web Channels
• Authentication
• Sharing and Permissions
• Environment Management
• Development, Test and Production Environments
• Power Platform Solutions
• Application Lifecycle Management (ALM)
• Production Monitoring
• Governance and Administration
• Enterprise Deployment Best Practices