Intents in Dialogue Tree as a Task
Updated
An Intent represents the purpose or goal behind a user’s input in a conversation. For example:
If a user says, “I want to book an appointment”, the intent is Book Appointment.
If a user says, “Cancel my order”, the intent is Cancel Order.
In AI Agent Studio, Intents are the foundation for understanding user requests, triggering the correct response or actions, and designing Dialogue Tree Tasks, which are structured, rule-based conversational flows often required in compliance-driven environments.
Intents act as building blocks that allow your AI Agent to interpret input and respond accurately. They are important because they enable deterministic branching in conversations, ensure compliance by mapping user input to predefined flows, and improve user experience by making interactions predictable and reliable.
Configure an ML Model
Before creating intents, you need an ML Model that can detect them. Here’s how to configure it:
1. Navigate to ML Model Configuration
Navigate to AI Agent Studio from the Sprinklr launchpad.
Click the Expand icon corresponding to the Build section.
Navigate to Tasks and click View.
On the Tasks Manager window, click the ML Models button on the open ML Models record manager.
On the ML Models record manager, click the Add ML Model button on the top right corner.
2. Create New Model
On the Create New Model screen, configure the following fields:
Field | Description |
Name | Enter a unique name for your model (e.g., Customer Support Model). |
Model Type | Select Zero-Shot Model. This type predicts intents without needing extensive training data. |
Priority | Numeric value that determines execution order when multiple models exist. Lower number = higher priority. |
Segmentation Regex (Optional) | Add a custom regex to split user input into smaller segments for better analysis. |
Share with other Applications (Optional) | Select the other applications from the dropdown if you want this model available across multiple apps. |
3. Save the Model
Click the Create button in the bottom right corner to save your ML Model. Once saved, you will be redirected to the Intent Record Manager for the respective model.
Create Intents in the ML Model
Follow the steps below to create Intents:
Click the + Intent button at the top right corner on the Intent Record Manager. The Intent Details screen will open.
Enter the Intent Details inputs as mentioned in the table below.
Field | Description |
Intent Model | Prefilled with the ML Model name you created (e.g., Shivi Test). |
Intent Name | Enter a clear, action-oriented name (e.g., Book Appointment). |
Description | Briefly describe what this intent does (e.g., Handles appointment booking requests). |
Confidence Threshold | Default is 0.3. This is the minimum confidence score required for the model to classify input as this intent. Lower threshold = more matches, less accuracy. |
Intent Priority | Lower numeric value = higher priority. Used when multiple intents are detected simultaneously. |
Click the 'Save' button the bottom right corner to save your Intent. Repeat this process for all intents you need.
Use Intents in Dialogue Tree Tasks
Once your intents are created:
Navigate to Dialogue Tree Tasks.
Map each intent to a specific branch in your conversation flow.
This ensures that when a user input matches an intent, the system follows the correct path.