Boost Email Efficiency with Power Platform AI Builder and OpenAI

Table of Contents

• Introduction
• Understanding AI Builder
• Building a Flow to Classify Emails
  • Scenario: Classification of Emails
  • Creating and Training the AI Model
  • Testing the Prompt
• Designing the Power Automate Flow
  • Initializing the Flow
  • Monitoring the Mailbox
  • Feeding Emails into the Prompt
  • Creating Conditions and Actions
  • Testing the Flow
• Conclusion

Building a Flow to Classify Emails with AI Builder

As businesses Continue to receive an influx of emails, it becomes crucial to efficiently categorize and prioritize them. This is where AI Builder comes in. AI Builder is a powerful tool that allows users to Create AI models to automate various tasks. In this article, we will explore how to build a flow using AI Builder to classify emails and take appropriate actions Based on their category.

Introduction

In today's digital world, businesses receive numerous emails on a daily basis. It can be a tedious task for staff to manually sort through these emails and assign them to different teams based on their priority. However, with AI Builder, businesses can automate this process and save valuable time and resources.

Understanding AI Builder

AI Builder is a Microsoft Power Platform tool that allows users to build AI models without the need for extensive coding knowledge. It provides pre-built models for various tasks, including classification and extraction, but also allows users to train their own models using custom data. In this article, we will focus on using AI Builder's GPT (Generative Pre-trained Transformer) model to classify emails.

Building a Flow to Classify Emails

Scenario: Classification of Emails

Let's imagine a scenario where You are running a company with a custom service mailbox. Your staff is responsible for reviewing and responding to the emails in this mailbox. To streamline the process, you want to automatically classify the incoming emails into different categories based on their content. You want to divide them into three buckets: "Archive" (no follow-up required), "Action" (customer asking for help or follow-up), and "Urgent" (customer complaining or requesting priority). Additionally, for urgent emails, you want to send a notification to your team's Channel for immediate Attention.

Creating and Training the AI Model

To start building the flow, we will use AI Builder's GPT model. In AI Builder, you can find various pre-built classification models, but for this scenario, we want to use the raw GPT model to have more flexibility. The GPT model uses Azure open AI models, so you need to set up your own Azure open AI models. After selecting the GPT model, create a prompt that defines the classification task. In this case, our prompt asks the model to classify the email message into three categories: archive, action, and urgent. We provide some sample emails to train the model, including an email with no follow-up required, an email asking for help, an email asking for urgent help, and an email complaining about the service. By training the model on these examples, it will learn to classify incoming emails accurately.

Testing the Prompt

Once we have defined the prompt and trained the model, it's essential to test its accuracy. We can use the prompt to classify sample emails and see if the model correctly categorizes them. By providing sample emails that match the categories we defined, we can evaluate the model's performance. For example, if we provide an email that requires no follow-up, the model should classify it as "archive." Similarly, an email asking for help should be classified as "action." By testing multiple examples, we can ensure that the model is accurately classifying emails based on their content.

Designing the Power Automate Flow

Now that we have created and tested the AI model using AI Builder, it's time to design the Power Automate flow to automate the email classification process.

Initializing the Flow

To initiate the flow, we will create a trigger that monitors the mailbox for new emails. In Power Automate, we can use the "When a new email arrives (V3)" trigger from the Outlook connector. This trigger will kickstart the flow whenever a new email is received in the specified mailbox.

Monitoring the Mailbox

Once the trigger is set, we can feed the email body into the prompt created using AI Builder's GPT model. This step involves using Power Automate's dynamic content feature to extract the email body and pass it to the prompt. By doing this, we ensure that the email content is classified accurately by the AI model.

Creating Conditions and Actions

After feeding the email into the prompt, we need to create conditions based on the classification result. We will use Power Automate's control actions, such as "Switch" and "Condition," to evaluate the output of the AI model and direct the flow accordingly. For example, if the email is classified as "urgent," we want to send a notification to the team's channel for immediate attention. We can achieve this by using Power Automate's "Post a message" action in the Teams connector. Similarly, for emails classified as "action," we can move them to a specific folder in the mailbox for the corresponding team to handle. For emails classified as "archive," we can simply move them to an archive folder.

Testing the Flow

To ensure the flow is working correctly, we can test it using sample emails. By sending emails that match the predefined categories, we can verify if the flow accurately classifies and takes appropriate actions based on the email content. It's essential to test various scenarios to ensure the flow handles different types of emails correctly.

Conclusion

In today's fast-paced business environment, automating mundane tasks like email classification can significantly improve efficiency and productivity. By leveraging the power of AI Builder and Power Automate, businesses can streamline their email management processes and provide faster response times to customer inquiries. The ability to accurately classify and prioritize emails using AI models not only saves time but also ensures that critical issues are addressed promptly. Start using AI Builder and Power Automate today to take AdVantage of these automation capabilities and enhance your email management system.

Highlights

• Automate email classification using AI Builder and Power Automate
• Create a flow that classifies emails into three categories: archive, action, and urgent
• Train an AI model using AI Builder's GPT model and Prompts
• Test the model's accuracy with sample emails
• Design a Power Automate flow to monitor the mailbox and take appropriate actions based on email classification
• Use Power Automate's control actions to create conditions and direct the flow
• Enhance efficiency and productivity by automating email management processes

FAQ

Q: Can I use AI Builder's GPT model for other tasks besides email classification? A: Yes, AI Builder's GPT model can be used for various text-based tasks, such as chatbot development, sentiment analysis, and content generation.

Q: How accurate is the GPT model in classifying emails? A: The accuracy of the GPT model depends on the quality and diversity of the training data. By providing a well-labeled dataset and testing the prompt with different examples, you can improve the model's accuracy.

Q: Is it necessary to have a human approval step in the flow for email classification? A: While it's recommended to have a human approval step in AI-related actions, it's not necessary for testing purposes or in scenarios where the model's accuracy is sufficient. However, for critical tasks or sensitive information, a human approval step can provide an extra layer of assurance.

Q: Can the email classification flow be extended to other platforms, such as social media? A: Yes, the email classification flow can be extended to monitor social media feeds or other platforms. By integrating APIs and connectors, businesses can automate the process of monitoring and categorizing content from various sources.