How to Improve Copilot Memory in Microsoft Copilot Studio (Setup Guide)

In this tutorial, we’ll guide you through enhancing your Microsoft Copilot’s memory capabilities using Copilot Studio. By the end of this setup guide, you’ll enable your Copilot to remember past conversations and user details, preventing repetitive interactions, and creating a more seamless and engaging user experience. Our step-by-step approach demystifies Copilot memory, making your interactive agent more intuitive and human-like.

What You’ll Learn

• Setting up short-term and long-term memory using variables
• Differentiating between topic and global variables
• Utilizing topic inputs and knowledge sources
• Implementing conditional logic to enhance agent interactions

Why This Matters

In the realm of generative AI, creating a seamless user experience is crucial. Copilot agents that constantly repeat information or require users to do so face decreased engagement. By mastering memory settings, your Copilot becomes a more useful and appealing tool, retaining user engagement and improving interaction quality.

---

Step-by-Step Instructions

Understanding Variables

Variables form the backbone of your Copilot’s short-term memory. They store information to manage what your agent remembers from one interaction to the next.

• Topic Variables are specific to a particular part or topic within the agent.
• Global Variables span across the entire agent, applicable in any context.

To manage these:

1. Navigate to your topic.
2. Click on Variables to view and manage topic variables.
3. Consider using global variables for information that needs to be accessible across different topics.

Using Topic Inputs

Topic inputs are enhanced variable versions that prevent redundancy. When a question is posed in the conversation pane, a new variable is created.

• To manage inputs, go to Details in your topic, then select Input to define variables.
• These inputs leverage generative AI to streamline conversations, ensuring the user does not have to repeat information.

Adding Knowledge Sources

Think of knowledge sources as your Copilot’s long-term memory. They provide information that isn’t needed immediately.

• Connect various knowledge sources such as PDFs, SharePoint sites, and websites.
• To set up a knowledge source, navigate to the Knowledge tab.
• Use generative answers nodes within your topics to direct when to search these sources.

Implementing Conditional Logic

Conditional logic helps your agent make decisions based on specific criteria, much like Power Automate flows:

1. Pose a question within your topic as a boolean field (e.g., “Do you need help writing an email?”).
2. Build logical branches: If “Yes,” execute certain actions; if “No,” proceed differently.
3. Employ conditions for various scenarios, such as when a variable is blank or a Power Automate flow fails.

Tips and Best Practices

• Lean on generative AI capabilities instead of overloading the canvas with conditional logic.
• Be mindful of using conditional logic to protect against potential pitfalls, like incomplete variable data or failed flows.
• Check out some Microsoft documentation here: https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-select-agent-model

---

Conclusion

Enhancing the memory of your Microsoft Copilot is about effectively managing variables, inputs, and utilizing your knowledge base strategically. By doing so, you ensure a smoother, more engaged interaction for users. Explore further resources to deepen your understanding and proficiency in Microsoft Copilot Studio, ensuring your Copilot remains a valuable asset in transforming user experiences.

For more insights and tutorials, check out the playlist on Microsoft Copilot Studio available on the Citizen Developer channel.