Documentation
Build
Information extractor

Information Extractor

Information Extractor allows your assistant to pull structured data from the conversation as it happens. Instead of just transcribing the call, the assistant identifies specific values like preferences, budgets, or intent and stores them as variables for use in webhooks, SMS, or CRM logging.

Extractor Types

You can choose from three types of extractors depending on the kind of data you need to collect:

TypeDescriptionBest For
Yes/No QuestionReturns a Boolean value (true or false).Checking interest, consent, or status.
Single ChoiceRestricts the AI to choose from a predefined list of options.Lead qualifying (e.g., "Hot", "Warm", "Cold").
Open QuestionAllows the AI to find the answer based on provided examples.Names, specific dates, or complex descriptions.

Creating an Extractor

Select Type

Go to Actions → Information Extractor, click + Create, and select the extractor type that fits your data needs.

Set the Identifier

Enter a unique variable name using lowercase letters and underscores (e.g., lead_interest or preferred_time).

This Identifier is the key used in webhooks and SMS templates (e.g., {identifier}).

Describe the Extraction Goal

In the What the AI should extract from the call field, provide a clear prompt for the AI.

  • Example: "Is the user interested in buying our services?"

Configure Specific Logic

Depending on the type selected, add your constraints:

  • For Single Choice: Click + Add Choice to define the only acceptable values the AI can pick.
  • For Open Question: Click + Add Example to show the AI the format or type of data you expect (e.g., "User wants to move in by October").

Save

Click Save. The extractor will now monitor all active calls and populate its value as soon as the information is detected.


How it appears in UI

Yes/No Extractor

Simple binary logic. If the AI detects a positive sentiment regarding the prompt, the identifier is set to true.

Single Choice Extractor

Ensures data stays clean for your CRM. You define the bucket (e.g., Morning, Afternoon, Evening) and the AI categorizes the user's response.

Open Question Extractor

The most flexible type. By providing examples, you guide the AI's "reasoning" to extract specific text from natural conversation.


Using Extracted Values

Once data is extracted, it can be utilized immediately:

  • Webhooks: The response is automatically included in the call webhook sent to your server.
  • Personalization: Use the identifier in a Send SMS action to confirm details back to the user (e.g., "Hi, we noted you are looking for a {bedroom_count} bedroom house.").
⚠️

Important: Only use lowercase letters and underscores for Identifiers. Avoid spaces or special characters to ensure compatibility with API integrations.