All Copilot courses

½ day Copilot series

Building Copilot Agents

Plan, Build and Deploy Copilot Agents

½ day (3 hours) Live online — English or French Intermediate Approximately 75% hands-on practice
Register

C$249 per workshop, per person

Course overview

Copilot Agents allow you to automate recurring tasks and create specialized assistants grounded in your organization's knowledge. This focused, hands-on workshop takes participants through the full agent lifecycle: plan, build, ground, test, refine and share.

Participants learn how to identify strong agent use cases, plan an agent using the Agent Design Canvas, build and configure the agent in Agent Builder, write effective instructions, connect knowledge sources such as SharePoint and OneDrive, test and refine responses, and understand governance and sharing considerations.

The session applies the Goal, Context, Constraints and Sources prompt framework throughout, with approximately 75% of the time spent on practical exercises using realistic business scenarios.

Course agenda

Unit 1 — Understanding and Planning Effective Agents

25 minutes

  • Understand what a Copilot Agent is
  • Distinguish general Copilot from specialized agents
  • Identify strong and weak agent use cases
  • Explore common business applications
  • Determine when not to create an agent
  • Compare Agent Builder and Copilot Studio
  • Understand the agent lifecycle: Plan → Build → Ground → Test → Refine → Share
  • Define purpose, audience, scope, tasks, knowledge, behaviour and boundaries

Hands-on practice: Select a business use case and create a concise Agent Design Canvas.

Unit 2 — Build and Configure the Agent

40 minutes

  • Access Agent Builder
  • Create an agent conversationally
  • Create from a template where appropriate
  • Work with agent configuration
  • Configure name, description, icon, instructions, knowledge, capabilities and starter prompts
  • Compare generated configuration with the original plan

Hands-on practice: Build the first functional version of the agent planned in Unit 1.

Unit 3 — Design Effective Agent Instructions

40 minutes

  • Define the agent's purpose
  • Create behavioural guidelines
  • Define skills
  • Establish step-by-step workflows
  • Specify tone
  • Establish response requirements
  • Set restrictions and boundaries
  • Handle missing information
  • Define error behaviour
  • Use examples effectively
  • Avoid conflicting instructions
  • Separate instructions from knowledge

Hands-on practice: Progressively rewrite and improve the agent's instructions, testing changes as you work.

Break

10 minutes

Unit 4 — Ground the Agent with Knowledge

35 minutes

  • Understand why grounding matters
  • Select appropriate knowledge sources
  • Work with SharePoint
  • Work with OneDrive and files
  • Work with websites and other supported sources
  • Use organizational information where available
  • Scope knowledge appropriately
  • Understand permissions
  • Distinguish knowledge from instructions
  • Reduce irrelevant responses through better source selection

Hands-on practice: Add appropriate knowledge to the existing agent and compare responses before and after grounding.

Unit 5 — Test, Troubleshoot and Refine the Agent

25 minutes

  • Use the testing experience
  • Create realistic test prompts
  • Test expected scenarios
  • Test edge cases
  • Test missing-information scenarios
  • Evaluate accuracy
  • Evaluate consistency
  • Evaluate adherence to instructions
  • Diagnose whether problems originate from instructions, knowledge, scope or user prompts
  • Refine instructions and sources
  • Improve starter prompts

Hands-on practice: Test the agent developed throughout the workshop and complete a structured refinement cycle.

Unit 6 — Sharing, Governance and Next Steps

5 minutes

  • Understand sharing considerations
  • Understand permissions
  • Maintain and update an agent
  • Recognize when Agent Builder is sufficient
  • Recognize when a solution should move to Copilot Studio
  • Identify practical next steps for building additional agents

Key takeaway

Participants leave with a practical methodology for planning, building, grounding, testing and refining Copilot Agents—and a clear understanding of when Agent Builder is the right tool versus when to move to Copilot Studio.

What we cover

  1. 1Understanding and planning effective agents
  2. 2Build and configure the agent
  3. 3Design effective agent instructions
  4. 4Ground the agent with knowledge
  5. 5Test, troubleshoot and refine the agent
  6. 6Sharing, governance and next steps