Skip to main content
The information extractor tells your agent which data to capture from the conversation: a first name, an order number, an address, a date. At the end of each call, the extracted values are saved with the call and can be reused in your tools. You configure it in the agent, from Information extractor in the left menu.
Information extractor page of an agent, showing the Customer informations group and the Independent information section

Two ways to organize information

You can create information inside a group or as independent information. The page shows one section per group, plus an Independent information section.

Information group

Gathers information that belongs together, such as a customer’s contact details. Each group has a name and can be renamed, deleted, or collapsed.

Independent information

A standalone piece of information with no link to the others, such as a case number or a call reason.
Use a group when the information describes a single item (a customer, a booking, a delivery). Use independent information when the data stands on its own.
Groups are top-level sections. A group contains information, not other groups.

Structure of an information field

Each field is displayed in a table with two columns: Field and Data type.

Create a group

1

Open the extractor

In your agent, click Information extractor in the left menu.
2

Add a group

Click Add at the top right, then choose the information group.
3

Name the group

Give it a clear name, for example Customer informations.
4

Add information

In the group section, click Add new information, fill in the name, type, and description, then confirm.
To edit or delete a field, use the pencil and trash icons on the right of its row. The same icons in the group header let you rename or delete the group.
Deleting a group deletes all the information it contains.

Create independent information

There are two ways:
  • In the Independent information section, click Add information.
  • Click Add at the top right and choose independent information.
Then fill in the name, type, and description.

Configuration example

For an appointment-booking agent, you can define:
  • a Customer informations group with First name, Last name, Phone number, and Email;
  • independent information such as Call reason or Requested date.
This split keeps the data readable: contact details stay together, call-specific information stays separate.

Import and export as JSON

You can export the extractor configuration as JSON, then import it into another agent. This is useful to:
  • reuse the same set of information across several agents;
  • back up a configuration before changing it;
  • prepare the list of information outside the interface, then load it in one go.
  1. Open Information extractor.
  2. Click the import/export icon to the left of the Add button.
  3. Choose export. The JSON file contains the groups and the independent information.

File format

Export an existing configuration first to get the exact format, then edit the file before importing it.

Write good descriptions

The description is what the model reads to decide what to extract.
  • State exactly what you expect: “Customer’s phone number, in international format” rather than “Phone”.
  • Specify the output format when it matters: date as DD/MM/YYYY, number without spaces.
  • Cover ambiguous cases: “If the caller gives two addresses, keep the delivery address.”

Publish your changes

Changes apply to calls only after you click Save, then Publish. While the agent has the Draft status, the previous configuration stays active.