Data extraction and outcomes

After a conversation ends, the platform can read its transcript and fill in fields you define, such as the caller's intent or a booking date. It can also judge whether the conversation reached the agent's goal.

Define extraction fields

1. Open the agent and choose Data Extraction.

2. Choose Add field.

3. Fill in the field:

   Setting	What it does
   Name	The field's key, for example user_intent. Use letters, numbers and underscores, and don't start with a number. Each name must be unique.
   Type	Text, Number or Yes / No.
   Description	Tells the extractor what to look for, for example "What the caller wants to achieve". Up to 500 characters.

4. Choose Save changes.

Write clear descriptions. The extractor relies on them to decide what goes into each field.

Example

Name	Type	Description
user_intent	Text	What the caller wants to achieve, in a few words.
appointment_date	Text	The date the caller booked, as YYYY-MM-DD. Empty if nothing was booked.
party_size	Number	How many people the booking is for.
wants_callback	Yes / No	Whether the caller asked us to call them back.

Where the results appear

To get the extracted fields, open Observe → Reports and export Extracted data for a date range. The export has one row per conversation.

Conversation goal and outcomes

Set a Conversation goal in the agent's Overview section, for example "Help the customer book an appointment". Once it is set, each finished conversation is judged against it, and the result shows as an outcome in Analytics.

Data extraction or memory?

• Use data extraction for fields you want to report on or export for every conversation.
• Use agent memory for facts the agent should recall the next time the same customer gets in touch.