Copilot Studio Tutorials

Not long ago, building an AI assistant meant hiring a developer, writing a lot of code, and waiting weeks before anything was ready to test. Microsoft Copilot Studio changed that completely.

Today, you can build a fully functional, AI-powered agent that answers employee questions, processes requests, connects to your SharePoint data, triggers Power Automate flows, and hands off to a human agent when needed — all without writing a single line of code. I’ve built these kinds of solutions for real organizations, and the results genuinely surprise people the first time they see it working.

The tutorials on this page cover everything I’ve learned building with Copilot Studio — from creating your very first topic to advanced generative AI configurations, custom integrations, and enterprise deployment. Written in plain language, step by step, with real examples throughout.

What Is Microsoft Copilot Studio?

Microsoft Copilot Studio is a low-code platform for building AI-powered conversational agents that can be deployed across Microsoft Teams, SharePoint, websites, and other channels.

It was formerly known as Power Virtual Agents, and Microsoft rebranded and significantly expanded it in 2023 as part of the broader Copilot wave across the Microsoft 365 ecosystem. If you’ve used Power Virtual Agents before, Copilot Studio will feel familiar — but with a lot more capability, especially around generative AI and the new agent framework.

With Copilot Studio, you can build agents that:

Answer questions based on your company’s SharePoint knowledge base, websites, or uploaded documents
Walk users through step-by-step processes (like submitting a leave request or raising an IT ticket)
Connect to backend systems through Power Automate flows or direct API calls
Integrate with Microsoft Teams so employees can interact with the agent right inside their daily workflow
Handle conversations using generative AI when no predefined topic matches the user’s question
Escalate to a live human agent when the conversation needs a real person
Act as an autonomous agent that can take multi-step actions on behalf of a user without constant prompting

It sits right at the intersection of Power Platform and Microsoft’s AI capabilities — making it one of the most exciting tools in the Microsoft 365 ecosystem right now.

Who Are These Tutorials For?

Complete beginners who have never built an agent and want to start from the very beginning
Power Platform developers who already work with Power Apps and Power Automate and want to add intelligent agents to the mix
SharePoint administrators who want to build agents that surface SharePoint content conversationally
IT and HR teams who want to reduce repetitive questions by building a self-service agent for employees
Business analysts who want to automate request and approval processes through a conversational interface
Intermediate Copilot Studio users who know the basics and want to go deeper into generative AI, custom entities, and enterprise deployment

I write every tutorial here assuming you’re new to Copilot Studio but not necessarily new to technology. I explain the concepts clearly before showing you how to build, so you understand what you’re doing — not just how to click through a wizard.

How Copilot Studio Works — The Core Concepts

Before you start building, let me give you the mental model that makes everything in Copilot Studio click.

A Copilot Studio agent is built around three core ideas:

1. Topics
A topic is a conversation path. It defines what the agent does when a user says something specific. Every topic has a trigger — a set of phrases that tell the agent “this is when you should activate this topic” — and a series of nodes that define what happens next: ask a question, show a message, call a flow, branch based on a condition, or end the conversation.

Think of topics as the building blocks of your agent’s knowledge and behavior.

2. Entities
Entities are the specific pieces of information your agent extracts from what the user says. If a user types “I need to book a meeting room for next Friday,” the agent needs to understand “meeting room” (what they want) and “next Friday” (when they want it). Entities capture those pieces. Copilot Studio has built-in entities for common things like dates, times, numbers, and email addresses — and you can create custom ones for your business-specific concepts.

3. Generative AI (Generative Answers)
This is the feature that really changed what Copilot Studio can do. Instead of only responding to topics you’ve explicitly defined, you can point your agent at a knowledge source — a SharePoint site, a set of documents, a website — and it will use generative AI to answer questions it wasn’t specifically programmed for. If the answer exists somewhere in your knowledge base, the agent finds it and responds naturally.

These three things — topics, entities, and generative AI — are the foundation of every agent you’ll build.

The Copilot Studio Interface — What You’re Looking At

When you open Copilot Studio at copilotstudio.microsoft.com, here’s what you’ll work with:

Agents panel — where you create and manage your different agents. You can have multiple agents in the same environment.

Topics — the section where you build and edit all your conversation paths. This is where you’ll spend most of your time.

Knowledge — where you add knowledge sources (SharePoint sites, documents, websites) that feed the generative AI capabilities of your agent.

Actions — where you connect your agent to Power Automate flows or other plugins that let it actually do things, not just answer questions.

Channels — where you configure where your agent will be deployed: Microsoft Teams, a SharePoint page, a custom website, and more.

Analytics — where you see how your agent is performing: how many conversations it’s had, which topics are working, where users are dropping off, and what questions it couldn’t answer.

What You’ll Learn in These Tutorials

I add new tutorials to this page regularly. Here’s what’s coming:

Getting Started

What is Copilot Studio and how does it fit into Microsoft 365
Setting up your first agent from scratch
Understanding the Copilot Studio interface
Creating your first topic and testing it in the built-in chat panel
The difference between system topics and custom topics
How Copilot Studio fits alongside Power Apps and Power Automate

Building Topics

Creating a topic from scratch — triggers, messages, and questions
Using question nodes to collect information from users
Branching conversations with condition nodes
Using the Message node — text, images, videos, and adaptive cards
Ending a conversation cleanly vs escalating to a human agent
Redirecting between topics — how to reuse common conversation paths
The fallback topic — what happens when the agent doesn’t understand

Entities and Slot Filling

Using built-in entities for dates, times, numbers, and emails
Creating custom entities with a fixed list of values
Creating regex entities for structured data like order numbers
Slot filling — how the agent collects multiple pieces of information naturally in a single conversation

Generative AI and Knowledge Sources

What generative answers are and how they work
Adding a SharePoint site as a knowledge source
Adding documents and files as a knowledge source
Adding public websites as a knowledge source
Configuring generative answers — controlling tone, response length, and content scope
Combining topics with generative answers in the same agent
Testing and tuning generative AI responses

Actions and Integrations

Calling a Power Automate flow from a Copilot Studio topic
Passing data from the conversation into a flow (variables to flow inputs)
Returning data from a flow back into the conversation
Using flows to read and write SharePoint list data
Using flows to send emails and Teams notifications
Calling HTTP endpoints directly from a topic (advanced)
Using connectors and plugins to extend your agent

Variables and Logic

What variables are in Copilot Studio and how they work
Topic variables vs global variables vs system variables
Setting and using variables across multiple topics
Using Power Fx expressions in Copilot Studio topics
Building conditional logic with multiple branches

Adaptive Cards

What adaptive cards are and why they make agent conversations better
Designing adaptive cards in the Adaptive Card Designer
Displaying information in a card format inside the conversation
Using adaptive cards to collect structured input from users
Sending adaptive cards through Teams

Channels and Deployment

Publishing your agent for the first time
Deploying to Microsoft Teams — step by step
Adding your agent to a SharePoint page
Embedding your agent on a public or internal website
Configuring authentication — who can interact with your agent
Using Microsoft Entra ID (Azure AD) authentication for internal agents

Analytics and Improvement

Reading the Copilot Studio analytics dashboard
Understanding session outcomes — resolved, escalated, abandoned
Identifying topics that aren’t working
Using conversation transcripts to improve your agent
A/B testing topic variations

Advanced Topics

Multi-language agents — building assistants that support more than one language
Copilot Studio and Dataverse — using Dataverse as a backend for your agent
Extending Microsoft 365 Copilot with Copilot Studio agents
Autonomous agents — building agents that take multi-step actions without constant user input
Governance and ALM — managing agents across development and production environments
Security and compliance considerations for enterprise agent deployment

Copilot Studio vs Power Virtual Agents — What Changed?

If you’ve worked with Power Virtual Agents before, you might be wondering what’s different. Here’s the honest answer:

The core canvas-based topic building is essentially the same — triggers, nodes, questions, conditions, and flow integrations all work similarly. What changed significantly is:

The terminology shift — Microsoft has fully moved away from the word “bot.” Everything is now called an agent. You build agents, publish agents, and manage agents. This reflects the broader shift in how Microsoft thinks about AI assistants — they’re not just reactive bots that answer questions, they’re proactive agents that can take actions
Generative AI is now a first-class feature — in Power Virtual Agents, AI responses were bolted on. In Copilot Studio, generative answers are central to how the product works
Knowledge sources — you can now point the agent at SharePoint sites, documents, and websites directly, without having to pre-program every possible answer
Power Fx support — you can now write Power Fx expressions inside topics for more sophisticated logic, which is huge if you already know Power Apps
Autonomous agent capabilities — agents can now be configured to take multi-step actions proactively, not just respond reactively to user messages
Microsoft 365 Copilot extensibility — Copilot Studio is now the tool you use to extend and customize Microsoft 365 Copilot with your own organizational knowledge and actions

If you know Power Virtual Agents, you’ll feel at home immediately. If you’re starting fresh, even better — you’re learning the current version without any old habits to unlearn.

Three Things That Will Save You Time Early On

1. Start with a focused, narrow use case
The biggest mistake people make with Copilot Studio is trying to build an agent that does everything for everyone on day one. Start with one specific use case — “answer common IT helpdesk questions” or “help employees submit leave requests” — and build it well. You can always expand later.

2. Add a knowledge source before you build every topic manually
Before you start hand-building topics for every question your agent should answer, add your SharePoint site or FAQ document as a knowledge source and enable generative answers. You’ll be surprised how many questions it handles correctly out of the box — which lets you focus your manual topic-building effort on the processes and workflows that genuinely need structured conversation paths.

3. Test constantly in the chat panel
Copilot Studio has a built-in test panel on the right side of the screen. Use it after every change. Don’t wait until everything is built to test — catch problems early when they’re easy to fix.

Here are the list of copilot studio tutorials.