Copilot Studio (Full) Basics
For simple agents with knowledge sources only, consider starting with Copilot Studio Lite. Use Copilot Studio Full when you need more control.
Creating Agents with Copilot Studio (Full)
Agent building in Copilot Studio gives you full control over conversational AI experiences with advanced capabilities like custom topics, Power Automate flows, and structured conversation logic.
Requirements
Microsoft 365 Copilot license
For more information on how to obtain a license:
What Makes Copilot Studio Full Different?
Copilot Studio Full offers more advanced capabilities than the M365 Copilot agent builder, known as Copilot Studio Lite:
Topics: Build structured conversation flows with nodes, conditions, and variables
Power Automate Integration: Use flows with connectors for complex automation
Premium Connectors are only available to Power Automate Premium Users, those with Pay as You Go on an Environment, or Credit Packs.
Advanced Publishing: Deploy to Teams and M365 Copilot Chat
Environment Management: Use Power Platform environments for collaboration within your unit
Enterprise Features: Agent orchestration, custom code, Azure AI Foundry models, advanced analytics
Getting Started
Creating Your Agent
Navigate to Copilot Studio and sign in. Select your environment from the top menu bar, then click Create and choose New agent.
There are two options:
Skip to Configure: This is for manual configuration where you can set the agent's name, description, instructions, and capabilities.
Describe It to Build It: Tell Copilot what you want your agent to do in plain language. It generates the foundation automatically, including initial topics and instructions.
Either way, you'll land in the agent authoring canvas. You will need to save the agent before doing some of the advanced configurations.
Understanding Topics
Each topic represents a portion of a dialogue and contains nodes that control the flow, sometimes based on context.
Topic Types
System Topics: Predefined topics that handle essential behaviors like ending conversations or escalating to a human. You can customize but not delete these.
Custom Topics: Topics you create from scratch or using AI to handle specific user requests, questions, or tasks.
Creating a Simple Topic
Go to Topics in the left navigation
Select Add a topic > From blank (or use Create from description with Copilot for AI assistance)
On the authoring canvas, you'll see a Trigger node
For Generative Orchestration (default for new agents and recommended): The trigger will show "The agent chooses". Give your topic a clear name and add a description of what the topic does. The AI orchestrator uses both the name and description to decide when to invoke the topic.
For Classic Orchestration: The trigger will show "User says a phrase". Add trigger phrases (keywords and questions users might ask that should trigger this topic).
Working with Nodes
Nodes are the actions your topic performs. Common node types:
Send a message: Display text, images, or adaptive cards to users
Ask a question: Collect user input and store it in a variable
Add a condition: Create if/then logic to branch conversations
Call an action: Trigger Power Automate flows or other external processes
Redirect to another topic: Hand off to a different topic
Variable management: Set, parse, or clear variable values
Generative answers: Pull responses from knowledge sources using AI
To add a node, click the + icon between existing nodes and choose the node type you need.
Note: Use the Copilot button in the toolbar to describe changes in plain language. It can add questions, modify messages, or create conditional branches for you.
A Simple Topic Example
Let's create an "Office Hours" topic for a UT department:
Trigger: Add description: "Provides office hours and contact information for UT department offices"
Ask a question: "Which office are you asking about?"
Save response to variable:
OfficeLocationMultiple choice options: Main Office, Student Services, Graduate Advising
Add a condition: Branch on
OfficeLocationvalueSend a message: Show hours for the selected office
Main Office: "Open Monday through Friday 8am to 5pm, (512) 471-0001"
Student Services: "Open Monday through Thursday 9am to 6pm, Friday 9am to 5pm, (512) 471-0002"
Graduate Advising: "Open by appointment only, email <fakeemail>@utexas.edu"
A note on variable naming: Name your variables with clear, descriptive names like OfficeLocation or UserEmail instead of generic names like Var1. It makes your topics much easier to understand and maintain later.
Always click Save in the upper right when you're done editing.
Sometimes the topic editor will display errors as you are building, this is normal. However once you’ve assigned all required values, it may still display an error. Click on Save, and confirm that you want to save with errors. This can sometimes (not always) resolve false positives.
Testing Your Agent
The Test pane appears on the right side of your screen. It updates in real-time as you build.
Type questions or statements to see how your agent responds
Check the Topics tab in the test pane to see which topics triggered
Click Reset to start a fresh conversation
Test thoroughly before publishing. Try different phrasings and edge cases.
Knowledge Sources
Add documents, SharePoint sites, or websites so your agent can answer questions using your organization's information.
Go to Knowledge in the left navigation
Click Add knowledge
Choose your source type:
Upload files: PDFs, Word docs, or text files
SharePoint: Link to SharePoint sites or document libraries
Public websites: Add URLs for the agent to crawl
Advanced options can include Microsoft Graph connectors for enterprise data sources, and foundry resources such as Azure AI Search. Most of the time these require advanced configuration and payment plans.
The agent uses AI to search these sources and generate conversational responses. This works alongside your topics. When a user asks something your topics don't handle, the Conversational Boosting system topic kicks in to provide generative answers.
Important: Users need proper permissions to the underlying SharePoint content. If they don't have access, the agent can't pull information for them.
Actions (Power Automate Flows)
Actions let your agent do things: send emails, create records, update databases, call APIs, and more.
Go to Actions in the left navigation
Click Add an action
Choose Create a new flow to build in Power Automate
Your flow must:
Include input variables if you're passing data from the conversation
End with a "Return value(s) to Power Virtual Agents" action if sending data back
Once created, you can call this action from any topic using a Call an action node. Map conversation variables to the flow's inputs and capture outputs back into variables.
Publishing & Channels
Publishing Your Agent
Go to Publish in the left navigation
Click Publish
Wait a few minutes for the process to complete
Publishing applies your latest changes to all connected channels. Users won't see updates until you publish.
Setting Up Channels
After your first publish, add channels to make your agent accessible:
Go to Channels in the top menu
Select a channel from the list:
Microsoft Teams: Share with colleagues, install in Teams app
Microsoft 365 Copilot Chat: Make available in M365 Copilot
Other channels are not available without additional cost, contracts, and approvals.
Each channel has its own configuration steps. Follow the channel-specific instructions.
Authentication: By default, agents use Microsoft Entra ID authentication for Teams and Microsoft 365 Copilot Chat.
Understanding Environments
Power Platform environments are containers for your agents, apps, flows, and data. Environments are good for collaboration within your unit and working with the AI Studio team.
If you need an environment set up for your team or project, reach out to AI Studio and they'll help get you configured. Further instructions to come.
Once you have an environment, you can switch between environments in Copilot Studio using the environment picker in the top menu.
Solutions
Agents in Copilot Studio can be added to Solutions, which are packages that help organize and move your work. Solutions allow you to export your agent and import it elsewhere if needed. More detailed documentation on Solutions will be provided later.
To add your agent to a solution:
Go to Solutions in Power Platform
Create a new solution or open an existing one
Add your agent to the solution
What to Avoid
Don't use Copilot Studio Full for:
Simple Q&A that M365 Copilot agents can handle
Tasks that should be automated without conversation (use Power Automate instead)
Scenarios requiring real-time critical uptime guarantees
When to use M365 Copilot Lite instead:
Simple knowledge-based agents
Agents that only need SharePoint or uploaded files as knowledge
Quick prototypes or personal productivity tools
Best Practices
Plan your topics carefully. Map out the conversation flows before building. Use flowcharts or diagrams.
Name your variables clearly. Use descriptive names like UserDepartment or RequestDate instead of generic names.
Test in a real channel. The test window doesn't perfectly represent Teams or M365 Copilot Chat.
Monitor analytics. Use the Analytics page to see which topics trigger, where users get stuck, and what questions go unanswered.
Secure your flows. Don't use premium connectors without proper authentication and data policies.
Document your agent. Maintain a guide explaining how the agent works, what topics exist, and how to modify it.
Key Differences from M365 Copilot Lite
Feature | M365 Copilot Lite | Copilot Studio Full |
|---|---|---|
Topics | No | Yes |
Premium Connectors | No | Yes* |
Environments | No | Yes |
Custom Code | No | Yes |
Advanced Analytics | Basic | Comprehensive |
Agent Orchestration | No | Yes |
Variable Control | No | Yes |
Getting Help
Microsoft Learn:
Official Microsoft Copilot Studio documentation - Microsoft Copilot Studio
Community Forums: Power Platform Community
UT Austin Support: Contact AI Studio