Introduction

The latest update introduces a powerful, AI-driven intent detection system integrated with OpenAI's ChatGPT, revolutionizing how chatbots interpret and respond to user inputs.

This system replaces traditional keyword-based detection with a dynamic, context-aware approach, enabling more natural and efficient interactions.

This summary provides a detailed, structured overview of the new features, setup procedures, and practical applications, ensuring users can leverage this technology to its fullest potential.

Deep Dive into the AI-Enhanced Intent Detection System

1. Activation and Setup

• Accessing the Feature:

  • Located under the Automations Tab within the Omni Channel Chatbot interface.

  • A dedicated Intents Section appears, initially blank, ready for configuration.

• Enabling Auto Intent Detection:

  • Requires connecting a verified OpenAI account.

  • Once activated, priority shifts:

    • Auto intent detection takes precedence over keyword matching.

    • The detection order becomes:

      1. Intents

      2. Keywords

      3. Default reply

• Prerequisites:

  • A verified OpenAI account linked to the workspace.

  • Proper API key configuration.

2. Creating and Managing Intents

• Methods of Intent Creation:

  • AI-Generated Intents: Use the "Try Generated by AI" feature for rapid setup.

  • Manual Trigger: Create intents with custom parameters for granular control.

• Naming and Language:

  • Intent names must be in English.

  • Chatbot responses can support any language supported by ChatGPT.

• Defining Parameters:

  • Essential for capturing user input details (e.g., date, time, name, number of guests).

  • Adding Parameters:

    • Mark as required.

    • Provide system messages for specific guidelines on data collection.

    • Attach parameters to custom fields or leave blank for JSON storage.

• Parameter Examples:

  Parameter Name	Required	Description	Custom Field	Values (Entities)
  Date	Yes	Reservation date	reservation_details.json	None
  Time	Yes	Reservation time	reservation_details.json	None
  Number of Guests	Yes	Party size	reservation_details.json	None
  Name	Yes	Guest's name	reservation_details.json	None

• Confidence Score:

  • Default threshold set at 0.8.

  • Ensures high accuracy in intent recognition.

3. Testing and Interaction

• Simulating User Input:

  • Use the Web Chat Widget to test intent detection.

  • Example interaction:

    • User: "I want to make a reservation."

    • Chatbot prompts for date, time, guests, and name sequentially.

  • Captured Data:

    • Stored as Unix timestamps for date/time.

    • Confidence scores typically above 0.95 indicate reliable detection.

• Handling Parameters:

  • Parameters are stored in JSON fields.

  • Can be mapped directly into flows for further processing.

  • Example:

    • Extracted date, time, name, and guests are assigned to variables for use in subsequent actions.

4. Managing Data and Flow Integration

• Clearing User Fields:

  • After capturing reservation details, clear custom fields to prepare for new interactions.

  • Ensures accuracy in subsequent bookings.

• Flow Configuration:

  • Use "Set Variable" actions to map JSON data into flow variables.

  • Example:

    • reservation_date = intent.reservation.date

    • guest_count = intent.reservation.guests

• Conditional Triggers:

  • If all required parameters are present, trigger the reservation flow.

  • If parameters are missing, trigger a follow-up flow to gather remaining data.

5. Generating Multiple Intents with AI

• "Generate by AI" Feature:

  • Creates a suite of related intents automatically.

  • Example for restaurant reservations:

    • Reserve Table

    • Cancel Reservation

    • Modify Reservation

    • Get Reservation Details

    • Search for Restaurants

    • Check Available Tables

• Advantages:

  • Saves time by avoiding manual intent creation.

  • Ensures comprehensive coverage of user intents related to reservations.

  • Adjust system messages and parameters post-generation for customization.

6. Practical Applications and Benefits

• Streamlined Reservation System:

  • Automates the entire booking process with minimal setup.

  • Handles multiple languages seamlessly.

  • Supports complex interactions with multiple parameters.

• Time and Resource Efficiency:

  • Reduces manual effort in defining keywords and intents.

  • Rapid deployment—intents can be generated in about 30 seconds.

  • Easily scalable for various use cases beyond reservations, such as support tickets, product inquiries, etc.

• Enhanced User Experience:

  • Natural language understanding leads to more human-like interactions.

  • Fewer misunderstandings due to context-aware detection.

  • Dynamic parameter collection improves response accuracy.

7. Best Practices and Tips

• Parameter Management:

  • Always mark essential parameters as required.

  • Use system messages to guide user input.

  • Clear custom fields after each interaction to prevent data leakage.

• Flow Design:

  • Incorporate conditional logic to handle incomplete data.

  • Use fallback flows for unrecognized inputs or missing parameters.

  • Map JSON data into flow variables for custom processing.

• Language Support:

  • While intent detection is in English, responses can be in any supported language.

  • Ensure parameters are clearly defined in English for consistent detection.

• AI-Generated Intents:

  • Regularly update and regenerate intents to adapt to evolving user needs.

  • Review generated intents for accuracy and relevance.