Someone in operations built the agent on a Friday afternoon. It answers questions about delivery status, the team uses it in Teams, and everyone is happy. Then someone asks who maintains it, what data it reads, how much it consumes, and what happens if it gives a carrier the wrong answer on Monday. Nobody knows.
If you're looking for a Copilot Studio course for your team, you're probably at that point. Building agents in Copilot Studio is easy to start; the hard part comes later. At Disrupsoft Academy we have five short courses on the tool, and we designed them as a path rather than a set of standalone talks. Here's what each stage covers, the order that works best, and what has changed on the platform since the courses were designed.
If you're still deciding whether Copilot Studio fits your needs, start with Copilot Studio: agents for your business without coding. This article assumes that decision is made and the question now is how to train the people who will build and maintain the agents.
Why a path instead of a single course
A test agent and a production agent have little in common. The test one does well in the demo. The production one has to connect to real data, respect permissions, have a predictable cost, and get better with use. Each of those jumps needs different skills, and cramming them into one session usually ends in a workshop where people see a lot and practice very little.
That's why our Copilot Studio training is split into five 2-hour courses, live and online, beginner level, each with a Disrupsoft certificate. You can take them separately, but they work better in order.
| Stage | Course | What it solves |
|---|---|---|
| 1 | Introduction to Microsoft Copilot Studio and Azure AI Studio | Understanding the tool and building a first agent |
| 2 | Design and Configuration of Intelligent Assistants | Getting the agent to understand what it's asked |
| 3 | Copilot Studio and Azure AI Studio: Enterprise Integration | Connecting the agent to Power Apps, Dynamics 365 and Dataverse |
| 4 | Advanced Conversational AI and Applied Machine Learning | Bringing the agent into Teams and analyzing its conversations |
| 5 | Deploying and Improving Virtual Assistants in Production | Licensing, monitoring and improvement once it's live |
Stage 1: understand the tool and build a first agent
Introduction to Microsoft Copilot Studio and Azure AI Studio is the starting point. It covers the basic structure of an agent: topics, logical flows and automated actions. The examples come from customer service, internal support and administrative tasks, which is where most companies try their first agent.
A note on the name. What the course calls Azure AI Studio is now Microsoft Foundry. Microsoft renamed it Azure AI Foundry first and then Microsoft Foundry, as noted in its guide to moving from the classic portal. It's the same platform, so if your tenant says "Foundry," that's what the course refers to.
Who should take it? Anyone who will build agents or ask for them, including department heads who need to know what's reasonable to request.
Stage 2: get the agent to understand the question
Design and Configuration of Intelligent Assistants starts from Copilot Chat and moves on to custom agents. The core of it is entities, regular expressions and prompts, the pieces that let an agent recognize an order number or a tracking code instead of guessing.
The cases come from logistics and distribution: carrier support, delivery orders, shipment tracking and inventory control. If you're not in logistics, it still applies. The underlying problem, capturing a piece of data correctly inside a conversation, is the same in sales or HR.
In our experience this is the stage people skip most, and the one that shows most later. An agent that misreads the question won't be fixed by loading more documents into it.
Stage 3: connect the agent to your systems
An agent that only reads a policy PDF can only go so far. It gets useful when it looks up and records information in the company's systems. Copilot Studio and Azure AI Studio: Enterprise Integration covers integration with Power Apps, Dynamics 365 and Dataverse, sending automated responses to external service systems, and using AI Builder to analyze trends in your data. It closes with a full integration case study.
There's a licensing change worth knowing before you get here. Starting November 1, 2026, new or renewed Power Platform and Dynamics 365 licenses no longer include seeded AI Builder credits, and those features draw on Copilot Credits instead, according to the AI Builder licensing documentation. Inside Copilot Studio, AI Builder already consumed Copilot Credits. The features are still there; what changes is which pool pays for them.
If your agent has to talk to in-house systems and not just Microsoft ones, we go over the architecture options in How to integrate AI agents into your company's software.
Stage 4: Teams and what the conversations tell you
Advanced Conversational AI and Applied Machine Learning shifts the focus to where the agent lives and what you learn from its conversations. It works with Microsoft Teams and Copilot, sentiment analysis applied to conversations on collaboration platforms, and agents that improve as people use them.
Teams matters for a practical reason: it's where people already work. An agent you have to go find on some other page gets used the first week and then forgotten.
Stage 5: production, costs and continuous improvement
Deploying and Improving Virtual Assistants in Production is the course that answers the questions from that Friday. It covers deployment in Copilot Studio, licensing models, built-in analytics to measure performance, session types and the metrics that matter, plus responsible AI and security. It ends with Foundry, for teams that need more control over models and scaling.
On licensing, here's the essential part from the Copilot Studio billing documentation:
- Agent usage is measured in Copilot Credits, and how many are consumed depends on the complexity of each response or action.
- There's pay-as-you-go: you link an environment to an Azure subscription through a billing policy in the Power Platform admin center and pay for what was used at the end of the month.
- Users with a Microsoft 365 Copilot license get Copilot Studio agent capabilities included when they use them inside Microsoft Copilot, Teams or SharePoint.
For monitoring, Copilot Studio analytics shows sessions, user questions, use of knowledge sources and tools, conversation outcomes and satisfaction. AI metrics such as sentiment analysis (still in preview) or theme classification refresh once a day.
Skip this stage and the agent stays the same as on day one while the business changes. Take it, and someone reviews every week where it fails and fixes it.
Building the path around who's being trained
Not everyone needs all five courses.
- If you'll build and maintain agents, take all of them, in order.
- If you manage the platform or the IT budget, prioritize 1 and 5: what an agent is and what it costs to keep running.
- If you lead a department that will request agents, course 1 is enough to know what to ask for and what not to.
- If your team already has a pilot that stalled, jump to 3 or 5, depending on whether the problem is integration or operations.
One warning: the courses don't replace deciding who owns the agent. That's the company's call, and it's better made before the training than after.
Frequently asked questions
Do I need to know how to code to take a Copilot Studio course?
No. All five courses are beginner level, and Copilot Studio is a low-code tool. What helps most is knowing the process of the area where the agent will be used.
How long is the full path?
Five courses of 2 hours each, live and online, so 10 hours in total. Each course comes with a Disrupsoft certificate.
What license do I need to use the agents my team builds?
It depends on where they're published. If they're used inside Microsoft Copilot, Teams or SharePoint and users have Microsoft 365 Copilot, those capabilities are included. Otherwise usage is charged in Copilot Credits, through purchased capacity or pay-as-you-go on Azure.
Where to start
If your team already has a test agent and you want to take it to production with people ready to maintain it, check out the path at Disrupsoft Academy or get in touch and we'll put together a plan based on the roles in your company.



