How Autonomy works

The orchestrator

The Autonomy orchestrator coordinates end-to-end check reviews across a case. It:

Retrieves all case data and identifies the checks or controls that require a review.
Loads the AI policies configured for each check type.
Delegates each check to the appropriate specialized review agent.
Aggregates the outcomes and generates a decision report for every automated approval or rejection.

Specialized check review agents

Each check type is processed by a dedicated AI agent.

Check type	What its agent does
AML	Uses LLM-based reasoning to classify potential match hits, discards false positives, validates true positives, and applies policy-based routing to each hit.
Document	Reviews submitted documents and approves or rejects them, comparing extracted data with case data to flag discrepancies and highlight potential issues.
ID document	Verifies personal identity documents and cross-checks data points, detecting mismatches, formatting anomalies and authenticity concerns.
ID verification	Confirms identity by comparing declared information to document data, and evaluates inconsistencies, potential fraud indicators and authenticity signals.

Every decision made by an agent includes a full reasoning report, accessible from the case. For AML, that report explains the disposition of every hit, line by line.

AI policies

Each review agent runs under check-specific AI policies where you define your own rules and decision logic.

Find the AI policies in the Advanced features settings menu.