Creating an AI Agent with Microsoft Copilot Studio

Imagine a world where an AI does not rebel like Skynet from Terminator but instead dedicates itself tirelessly to handling mundane tasks, freeing you to focus on what truly matters. Thankfully, Copilot’s Agents are all about productivity, not world domination! Think of them as intelligent, adaptable, and friendly helpers completely under your control. 

Microsoft Copilot is revolutionizing how you work by allowing you to create AI-powered Agents capable of tackling tasks that might otherwise consume hours of your time. Whether it is managing emails, analyzing data, or guiding customer interactions, these Agents are not just tools of efficiency, they are engines of empowerment. 

The best part? Copilot’s Agents come without a sinister agenda, unlike certain robots with thick Austrian accents. They are here to assist, support, and enhance your workflows, letting you focus on creative problem-solving and strategic thinking.

In this guide, we will explore how to build and deploy just one of many possible Agent types. These types are designed to help users access specific information on a site without the hassle of navigating through multiple pages. Keeping you ahead of the curve does not require time travel.

An Agent is an AI-powered assistant created within Microsoft Copilot Studio, designed to perform specific tasks or provide support based on user needs. Built using the robust capabilities of Copilot Studio, Agents are tailored to answer common employee questions for businesses, streamline workflows, handle repetitive tasks, and enhance user experiences in general.

By connecting business logic with natural language capabilities, these Agents will bring the power of AI directly into your processes, making them more innovative and user-friendly.

Developing a virtual support agent to assist users in navigating websites is one of the most frequent requests UDS receives from clients. 

The need for such an Agent stems from many users' challenges when searching for specific information across complex or content-heavy sites. An effective virtual support Agent should provide intuitive assistance by understanding user queries and delivering accurate results without requiring users to browse numerous pages. This ability not only reinforces the user’s experience but also refines accessibility and shortens the time spent searching for information.

Before we proceed, let us assume you already have access to and the necessary license for Copilot Studio.  

To begin with, you have two options for creating an Agent: using a natural language prompt for a quick setup (1) and manually configuring all the required settings (2).  

Here, we will concentrate on the second variant: 

Step 1. Click on the Agents item in the left sidebar (2) to start.

Step 2. On the Agents page, click on the “+ New agent” button (3).

Step 3. While exploring prompts and reviewing their results is always a great idea, this tutorial will focus on the manual Agent creation approach, so click on the “Skip to configure” button (4) on the next page.

Step 4. On the next page, you will find several options to configure your Agent, listed from top to bottom:   

• Language: Select the language your Agent will use.  
• Name and Icon: Set up your Agent’s identity by giving it a name and electing an icon.
• Description and Behavior Instructions: Provide a clear description and define the behavior guidelines for your Agent. 
• Once you have completed these settings, scroll to the bottom of the page and click on the “Add Knowledge” button (5) to proceed.

Step 5. Because our Agent will be using a public website to retrieve the information, we will select the “Public website” option (6) on the next screen.

Step 6. Now, enter the public website name and click on the “Add” button (7).

Step 7. If the site was added successfully, click on the “Add” button (8) at the bottom of the form.

Step 8. And now, it is finally time to create an Agent, so click on the "Create" button (9) in the top right corner.

Step 9. After the Agent has been successfully created, a sidebar on the right (10) will open, allowing us to test it.

! Note. Before we move on to testing, take a moment to finetune your Agent for optimal performance. 
  
Step 10. To keep the Agent grounded in your specific knowledge source, disable the ‘General knowledge’ option (12) on the ‘Overview’ tab (11).

Step 11. On the” Topic” tab (13), open the “System” group (14) and select the ‘Conversation Start’ topic (15).

Step 12. Here, we can edit the "Conversation Start" message to make our Agent look more personalized. You can do this by transfiguring the text in the “Message” box (16). 

! Note. Do not forget to save your changes (17).

Step 13. Because the Agent should be available for every person visiting your website, we need to disable authentication for it. For this, click on the “Settings” button (18) at the top right.

Step 14. Now, you need to go to the "Security" section (19), open the "Authentication" item (20), choose the ‘No authentication’ option (21), save (22), and confirm your changes.

Step 15. The final step is to publish your latest changes. For this, close the "Settings" panel by clicking the X in the top-right corner of the page. Then, click the “Publish” button (23) next to “Settings” to make your updates live.

! Note. You will get a potential risk warning related to disabled authentication. Ignore it and click “Publish”.

Your Agent is now ready for proper testing:

• You can test it directly within the right sidebar of Copilot Studio, but if you want to demonstrate it to others, it is preferable to use the "Go to demo website" option (24). To do this, click on the three-dot icon to the right of the "Settings" button.

• A new webpage will open with your Agent's window embedded in it. Share the link to this webpage with the person or group you want to involve in your Agent testing.

• Now, let us ask the newborn Agent some questions about UDS!

As you can see, the Agent is fully equipped to assist the site visitors even at this early stage.

And there is plenty of room to intensify the Agent even more. For example, we could enable it to create support tickets directly from the chat or schedule calls with UDS. Anyway, these Agent features are topics for other UDS Blog updates.

Together, we have gone through the steps to create a functional and practical Copilot Agent, demonstrating that the process is simple with the right tools and guidance. Copilot Agents have immense potential to streamline workflows, enhance user experiences, and bring real value to your business.

At UDS, we have the experience and expertise to make the most of this technology. Whether you want to build a simple support Agent or something more advanced, we can guide you every step of the way or handle the entire process for you.