Skip to main content

AI in Risk

Purpose​

An example of applying AI Conveyor in the context of the risk department: typical initiatives (scoring, AML, stress tests, early warning), and the specifics of stage gates and governance.

Key ideas​

  • Initiatives are often related to scoring models, anomaly detection, default forecasting, and regulatory reporting.
  • Requirements for explainability, model validation, and audit are heightened; governance and stage gates account for alignment with validation and compliance.
  • Value is measured in risk metrics (PD, LGD, the number of cases identified), cost savings, and reduced regulatory risk.

How it works​

Ideas from the risk department are registered in the common portfolio and go through the conveyor. At the prototype and production stages, the involvement of validation and compliance with internal and regulatory requirements is mandatory. Playbooks are adapted to the need for documentation for the regulator and for model reproducibility.