Skip to content

Search is only available in production builds. Try building and previewing the site to test it out locally.

Customer Facts

Customer facts are individual pieces of structured information the AI agent collects and remembers about a contact — such as their name, email, or address. They build up automatically as conversations happen, so the agent doesn’t have to ask twice.

The fact catalog defines which facts the agent may collect. Each entry has a key (a dotted path like kyc.legalName), a kind (text, email, phone, number, boolean, address, and so on), optional labels, and an optional validator and AI prompt hint that guide how the value is captured and checked. Facts can be organized into groups to keep the catalog tidy.

You manage the catalog under AI Agents → Configure → Customer Facts.

Facts are gathered from several sources, each shown with a label on the contact’s record:

  • From customer — shared in a message.
  • From operator — logged by your team.
  • Imported — brought in from existing history.
  • Inferred — deduced by the agent (these may show a confidence score).

Collected facts appear on the contact’s profile and in the conversation side panel, grouped by catalog group. A fact that’s saved but missing from the catalog is flagged as Uncatalogued so you can add a definition for it.

  • Contacts — where collected facts are shown.
  • The AI Agent — collects facts during conversations.
  • Glossary — definitions of key terms.