Using AI Intent Detection in Workflows

Understanding AI Intent Detection

The AI Intent Detection action is a powerful workflow tool that uses artificial intelligence to evaluate the emotional tone of a piece of text. It automatically reads and interprets text from sources like messages, form entries, or survey answers, then sorts it into one of three categories based on the detected sentiment. This allows you to create automated, intelligent responses that are tailored to how your contacts are feeling.

How It Works

When the action runs, it analyzes the provided text and directs the workflow down one of three automatically created paths:

• POSITIVE: For text that expresses happiness, satisfaction, or approval.
• NEGATIVE: For text that conveys disappointment, frustration, or dissatisfaction.
• NONE: For text that is neutral, ambiguous, or where a clear sentiment cannot be determined.

Benefits of Using AI Intent Detection

• Automate Sentiment Analysis: Instantly understand customer emotions without manual reading.
• Build Smarter Workflows: The automatic branching simplifies creating different follow-up actions for positive, negative, and neutral feedback.
• Improve Response Relevance: Send appropriate, empathetic replies or offers based on the customer's mood.
• Scale Customer Care: Handle high volumes of feedback automatically, ensuring no sentiment goes unnoticed.

Configuring the AI Intent Detection Action

Step 1: Add the Action to Your Workflow

In your workflow builder, locate the action by searching for "AI Intent Detection" in the actions menu. Drag and drop it into your workflow at the stage where you want the sentiment analysis to occur.

Step 2: Configure the Action Settings

Once added, you need to configure two main settings:

• Action Name: Give the action a clear name for your reference, such as "Analyze Survey Sentiment."
• Input Text: This is the text the AI will analyze. You can enter a static piece of text, or more commonly, use a dynamic variable. Click the custom value picker to insert a variable like {{message.body}} for an incoming SMS, or any other variable that contains the text you want to classify.

Step 3: Build Your Branch Actions

After saving the configuration, you will see three distinct branches stemming from the action: POSITIVE, NEGATIVE, and NONE. You can now build out the unique automation for each path.

• For the POSITIVE Branch: Add actions like sending a thank-you message, requesting a review, offering a referral bonus, or adding a "Happy Customer" tag.
• For the NEGATIVE Branch: Add actions to escalate to support, send an apology with a discount code, schedule a callback, or add a "Needs Help" tag.
• For the NONE Branch: Add actions for neutral outcomes, such as sending a follow-up question for clarification, routing the contact to a default campaign, or flagging it for manual review.

Example Applications

Managing Customer Support Feedback

After a support ticket is closed, send a satisfaction survey. Use AI Intent Detection on the response:

• A POSITIVE reply like "The agent was fantastic!" could trigger an automated request for a Google review.
• A NEGATIVE reply like "My issue wasn't solved" could automatically create a new, high-priority ticket and notify a manager.
• A NONE or short reply like "It was fine" could trigger a simple thank-you email.

Following Up After a Purchase

After a product is delivered, send a check-in email. Analyze the customer's reply:

• POSITIVE sentiment could trigger a workflow to ask for a product testimonial.
• NEGATIVE sentiment could trigger a workflow with return instructions and a special offer.

Pricing and Usage

This is a premium action. Each execution of the AI Intent Detection action incurs a small fee.

Common Questions

How accurate is the sentiment detection?

The action uses advanced AI models designed for high accuracy in classifying text sentiment. For best results, ensure the text you provide is clear and complete.

Can I use this action multiple times in a single workflow?

Yes, you can use multiple AI Intent Detection actions in a workflow to analyze different pieces of text or to perform layered analysis.

What if a message has both positive and negative elements?

The AI evaluates the overall tone and dominant sentiment of the entire text. If the sentiment is truly mixed or unclear, the action will typically route the workflow through the NONE branch.

Can I test how it will categorize text?

Yes. Use the workflow's test mode with sample text inputs to see which branch the action follows before publishing your workflow.