Output variables are data extracted from the conversation after the call ends. For each variable you want to extract, you write a specific extraction prompt that tells the system how to pull that information.

​
Overview

Think of output variables as a way to automatically collect and structure data from your conversations. After each call, the system:
  1. Reads the entire conversation transcript
  2. For each variable, applies its extraction prompt to pull out the relevant information
  3. Returns the extracted data in a structured format you can use for analytics, CRM updates, follow-ups, or compliance
This is useful for turning unstructured conversations into structured data without manual transcription or analysis.

​
How it works

For each output variable, you define:
  1. Variable name — what you want to call it
  2. Extraction prompt — instructions on how to extract that specific piece of information from the conversation
Example variables:
Extraction prompt: “Based on the conversation, determine if the customer’s overall sentiment is positive, neutral, or negative. Consider their tone, word choice, and expressed concerns.”Returns: "positive" | "neutral" | "negative"
Extraction prompt: “Extract the loan amount the customer mentioned they want. Return as a number in rupees.”Returns: 500000 (numeric value)
Extraction prompt: “What method did the customer prefer for follow-up: phone, email, or SMS?”Returns: "phone" | "email" | "SMS"
Extraction prompt: “What should be the next step in the process? Schedule callback, send documents, process application, etc.”Returns: "schedule callback" | "send documents" | "process application"
After the call, you get back structured data: 
{
  "customer_sentiment": "positive",
  "loan_amount_requested": 500000,
  "preferred_contact_method": "email",
  "next_action": "send documents"
}

​
Use cases

  • Lead qualification — sentiment, interest level, budget, timeline, decision maker status
  • Feedback collection — satisfaction score, pain points, suggestions, NPS
  • Order management — product details, quantity, delivery address, special requests
  • Compliance — consent given, disclosures acknowledged, customer decisions

🖼️ Add image

Placeholder — add screenshot showing how to configure output variables and their extraction prompts here.