AI console

Every model call runs as a principal derived from your own permissions, and every tool call is audited — including the ones that get blocked.

AI operations →

What the assistant may do

Allow-listed tool families

  • Retrieve and explain customer profiles, transactions and ML scores
  • Search operator knowledge and draft support replies
  • Draft segment definitions, campaign copy and journey outlines
  • Run approved analytics queries

What it can never do

Blocked in code, not in the prompt

  • Credit or debit a wallet
  • Approve or reject a withdrawal
  • Override KYC, AML or fraud decisions
  • Remove a responsible-gambling control or lift a hard suppression
  • Suspend or close an account
  • Grant an uncapped bonus

Ask the assistant

Runs as a principal derived from your own permissions — it cannot read what you cannot

It answers, summarises and drafts. It cannot send, approve, credit, activate or change anything.
AI draftGenerated by configured instance model. Review before use — requires human approval.

Every answer above is produced by the configured model from material it was handed. Where it says NOT_FOUND, the knowledge base did not answer the question — that is not the same as there being no answer, and it is not an invitation to guess one.