Build prompt actions for your copilot

Introduction

In this tutorial, we will guide you through the process of building prompt actions for your custom Copilot using Microsoft Power Platform. We will utilize the custom prompt authoring feature of AI Builder to instruct the GPT model effectively. By the end of this tutorial, you will be able to create prompts, define settings, test them, and configure your Topics using enterprise data from Dataverse.

Step 1: Add a Topic

To start, you need to create a Topic that will serve as the foundation for your prompt.

1. Open Microsoft Copilot Studio.
2. Navigate to the Topics section.
3. Click on "Add a Topic."
4. Provide a descriptive name for your Topic to easily identify its purpose.

Step 2: Create a Prompt

Next, you'll create a prompt that instructs the GPT model on what to do.

1. Within your Topic, select the option to "Create a prompt."
2. Enter the specific instructions you want the model to follow.
3. Ensure your prompt is clear and concise for better performance.

Step 3: Add Input Parameter

To customize your prompt, you can add input parameters that allow for dynamic responses.

1. Find the section for input parameters within your prompt settings.
2. Click on "Add input parameter."
3. Define the type of input you expect (e.g., text, number).
4. Provide a brief description of what the input parameter represents.

Step 4: Add Data Used

Utilize the Preview feature to ground your prompt with enterprise data from Dataverse.

1. Look for the "Add data used" option in the prompt settings.
2. Select the relevant data source from your Dataverse.
3. Ensure the chosen data enhances the prompt's context and relevance.

Step 5: Define Output

Specify what kind of output you expect from the prompt.

1. Go to the output settings of your prompt.
2. Clearly define the expected output format (e.g., text, JSON).
3. This helps in ensuring that the model produces usable responses.

Step 6: Configure Settings

Adjust your prompt settings for optimal performance.

1. Access the settings section of your prompt.
2. Review options such as response length, temperature, and model selection.
3. Set configurations that best fit your use case.

Step 7: Build the Prompt with Input and Data

Combine your input parameters and data to create a comprehensive prompt.

1. Review your prompt structure.
2. Ensure that all input parameters and data are correctly integrated into the prompt.
3. Make adjustments as necessary to improve clarity and effectiveness.

Step 8: Test the Prompt

Before finalizing, it’s essential to test your prompt to see how well it performs.

1. Locate the testing section of Copilot Studio.
2. Input sample data based on your defined parameters.
3. Run the test and analyze the output generated by the model.
4. Make any necessary adjustments based on the test results.

Step 9: Add and Configure Prompt Node

Now that your prompt is tested, you can add it to your Copilot.

1. Click on "Add Prompt Node."
2. Configure the node settings to define how the prompt will function within your Copilot.
3. Ensure it aligns with your overall application structure.

Step 10: Add and Configure Message Node

Finally, integrate a message node to handle the output from your prompt.

1. Click on "Add Message Node."
2. Set it up to display or process the responses generated by your prompt.
3. Customize it to fit the user experience you want to create.

Step 11: Test Prompt in Copilot Studio

Conduct a final test of the complete setup to ensure everything works seamlessly.

1. Navigate to the testing interface in Copilot Studio.
2. Execute the prompt with various inputs to verify functionality.
3. Review the outputs for accuracy and relevance.

Step 12: Explore Power Platform Prompt Library

To enhance your prompts, explore the Power Platform Prompt library.

1. Visit the Power Platform Prompt library.
2. Filter prompts by AI Builder to find useful templates.
3. Consider using the "fundraiser-information" prompt as a reference.

Conclusion

You have successfully learned how to build prompt actions for your Copilot using Microsoft Power Platform. By following these steps, you can create effective prompts, integrate enterprise data, and enhance the capabilities of your custom Copilot. For further exploration, consider experimenting with different prompts in the Power Platform library to refine your skills and improve your applications.