How to Build a Multi-Step Conversational Copilot Agent Using Microsoft Copilot Studio

If you’ve been curious about how to create a personalized, contextual multi-step copilot agent that can follow a conversation flow and gather the exact information you need, you’re in the right place. In this tutorial, I’ll walk you through building a fully interactive conversation in Microsoft Copilot Studio. Whether you’re automating coffee orders or gathering customer info for a database, structuring your copilot’s dialogue is key — and I’ll show you how to do it step by step.

Why Build a Structured Multi-Step Conversation?

Typically, when working with generative AI, the goal is to keep conversations natural and flexible. However, there are scenarios where a structured dialogue is necessary. For example, if your copilot needs to pull data from a database using a custom connector, you’ll want to collect very specific inputs — like an account number, phone number, or first name — in a well-defined order.

By carefully designing your copilot’s questions, conditions, and variables, you can control the flow of the conversation so that users provide exactly the data your system requires. This method balances the power of AI with the precision of traditional input gathering.

---

Getting Started: Creating a New Topic in Copilot Studio

Let’s dive in. In Microsoft Copilot Studio, your copilot is made up of topics, which define distinct conversational areas or purposes. For this example, we’ll build a simple coffee ordering topic.

1. Create a New Topic
   Start with a blank slate by creating a new topic called “Order a Coffee.” This topic will trigger whenever someone wants to place a coffee order.
2. Set the Trigger
   The trigger tells the copilot when to activate this topic based on user input. For instance, if a user says “I want a coffee,” the copilot will know to launch this flow. The trigger itself is not a message but an instruction for when to use the topic.

---

Building Your Questions and Variables

Step 1: Ask What They Want to Order

Start the conversation by asking a multiple-choice question: “What would you like to order?” For our coffee shop, options are limited to three: Coffee, Espresso, or Tea.

• Using a multiple choice entity ensures users can only pick valid options, improving data quality.
• Each question generates a variable — in this case, we’ll call it orderType — that stores the user’s choice.

Step 2: Conditional Follow-Up Questions

Next, the conversation branches depending on what the user ordered:

• If orderType is Coffee, ask, “What sort of coffee would you like?” Options might include dark roast, blonde roast, or seasonal roast.
• If orderType is Espresso, ask, “How many shots of espresso would you like?” Here, you’d use a number input.
• For Tea, you might ask a different question or skip some entirely.

By adding conditions based on variables, you create a dynamic conversation path tailored to the user’s choices.

---

Advanced Flow Control: Combining Conditions and Reusing Questions

To avoid duplication, you can use logical operators in conditions. For example, a question like “How would you like your coffee or espresso?” can be shown if the order is either coffee or espresso.

You can also insert variables into your question text dynamically. For example:

“How would you like your [orderType]?”

This makes your conversation more natural and adaptable without needing separate questions for each item.

---

Capturing Customer Details and Order Summary

After gathering order details, it’s common to ask for the customer’s name. Copilot Studio provides a prebuilt ‘person’s name’ entity, making it easy to capture this information accurately.

Once you have all the info, you can send a summarized message back to the user. This is where adaptive cards shine:

• Adaptive cards are customizable, interactive UI components you can design in Copilot Studio.
• You can include all captured variables — like coffee type, shots, additions, and customer name — in a neat summary card.
• Adaptive cards make the conversation visually appealing and clear, enhancing user experience.

---

Integrating with Power Automate and Custom Connectors

Finally, you can connect your copilot to external systems such as:

• Power Automate flows, which can trigger backend processes like sending the order to a kitchen system or saving it to a database.
• Custom connectors, which allow you to interact with APIs directly from your copilot.

At release, this integration required Power Automate, but now you can also connect directly using custom connectors, offering more flexibility.

---

Wrapping Up

Building a multi-step copilot conversation in Microsoft Copilot Studio takes a bit of upfront planning but pays off by delivering structured, efficient, and user-friendly interactions. From creating topics and triggers, designing questions with variables, adding conditions for branching logic, to integrating with backend systems, Copilot Studio gives you a powerful platform to create tailored AI assistants.

If you want personalized help or have specific questions, feel free to reach out. I’m passionate about Microsoft Copilot Studio and love coaching others to harness its full potential.

Happy Developing!