Building Microsoft Copilot Agents (1 day)

This one-day hands-on workshop is designed for Microsoft 365 power users, IT professionals, and citizen developers who want to move beyond Copilot Chat and design production-ready Copilot agents. Participants will learn when to build an agent versus using Copilot Chat, Power Automate, or a traditional application, and where agents fit within the broader Microsoft Copilot extensibility model. Through a structured build progression, delegates will design and implement a knowledge-grounded declarative agent in the morning, then extend it in the afternoon with actions, automation, and autonomous behaviours. The course emphasises real-world enterprise considerations including governance, security, cost control, risk management, and responsible AI design. By the end of the workshop, participants will have built, tested, and deployed a practical Copilot agent aligned to a genuine organisational use case.

Objectives

Evaluate when to build a Copilot agent versus using Copilot Chat, Power Automate, or a traditional application
Explain what Microsoft Copilot agents are and how they extend Microsoft 365 Copilot
Distinguish between declarative agents, autonomous agents, and custom engine agents (architectural awareness)
Design and build a knowledge-grounded declarative agent using Agent Builder and Copilot Studio
Connect agents to organisational knowledge and external systems using connectors and APIs
Configure actions and automation using Power Automate and OpenAPI integrations
Design autonomous agent workflows with appropriate guardrails and human oversight
Apply governance, compliance, cost control, and responsible AI principles to agent deployment
Deploy, share, and manage agents securely within Microsoft 365

Prerequisites

Working familiarity with Microsoft 365 applications such as Teams, SharePoint, and Outlook
Experience using Microsoft Copilot Chat or Microsoft 365 Copilot in day-to-day work
Access to a Microsoft 365 tenant with appropriate Copilot and Copilot Studio licences enabled
No prior development experience is required, although familiarity with data concepts and basic automation workflows is helpful

Strategic Foundations: Where Copilot Agents Fit

Differentiate between Copilot Chat, Microsoft 365 Copilot, Power Automate, and Copilot agents
Evaluate when an agent is the right solution versus automation or application development
Describe declarative, autonomous, and custom engine agents at a conceptual level
Understand the Microsoft Copilot extensibility model and how agents integrate with Teams and Microsoft 365
Review enterprise use cases across HR, IT helpdesk, sales, finance, and operations

Designing a Production-Ready Declarative Agent

Define a clear business use case and success criteria for an organisational agent
Navigate the Microsoft 365 Copilot Agent Builder and Copilot Studio interfaces
Create a declarative agent with structured system instructions and persona definition
Apply prompt and instruction engineering techniques to improve reliability and control
Test and iteratively refine agent behaviour using realistic scenarios
Understand how agent manifests and the Teams app framework underpin deployment

Grounding Agents in Organisational Knowledge

Explain how grounding works and how Microsoft Graph security trimming protects data
Connect agents to SharePoint sites, document libraries, and OneDrive folders
Add websites and external content sources
Use Microsoft Graph connectors to integrate third-party systems
Evaluate knowledge quality, scope, and freshness impacts on output accuracy
Diagnose and reduce hallucinations in grounded responses

Extending Agents with Actions and Automation

Distinguish between informational responses and task-executing actions
Add Power Automate flows as agent actions
Configure REST API integrations using OpenAPI definitions
Use Microsoft 365 and Dataverse connectors to read and write organisational data
Design safe input validation, confirmations, and user control mechanisms
Test actions end-to-end within Copilot Studio and Teams

Designing Autonomous Agents with Guardrails

Identify suitable use cases for autonomous agents
Configure triggers such as schedules, events, and inbound communications
Design multi-step reasoning and tool-use workflows
Implement human-in-the-loop approval and escalation patterns
Assess cost implications, token consumption, and runaway loop risks
Monitor agent runs and analyse outcomes using Copilot Studio analytics

Deploying, Managing, and Governing Agents

Publish agents to Teams, SharePoint, and Microsoft 365 Copilot
Manage versioning, updates, and rollback
Control agent availability through Microsoft 365 and Teams admin centres
Apply data loss prevention, sensitivity labels, and compliance controls
Use audit logs and Microsoft Purview to monitor activity
Define organisational policies for agent lifecycle management and change control