Default Risk Prediction Agent

Purpose

The Default Risk Prediction Agent is designed to analyze claims and disputes to identify patterns, risks, and corrective actions. It supports banking-related use cases by turning text and table-based inputs into usable summaries and structured analytics.

Primary users

The primary users are external, client-facing users. The agent is associated with the CBA team and is owned by Geoffroy Borensztejn.

Where it fits (process/stage/trigger)

The agent fits into processes where claims and disputes need to be reviewed, analyzed, and converted into insights. It can be used when text or table-based information about claims and disputes is available for analysis.

Key capabilities / workflow

The agent analyzes claims and disputes, checks whether the provided inputs are sufficient, extracts patterns and risks, identifies whether corrective actions are needed, and generates text summaries and structured analytics. If the analysis needs refinement, the workflow can loop back to review the inputs again.

Inputs

Typical inputs are text and tables related to claims and disputes. No created dataset is specified for this agent.

Outputs / Deliverables

The expected outputs are text summaries and structured analytics. These deliverables highlight identified patterns, risks, and corrective actions based on the provided claims and dispute information.

Value

The agent helps transform claims and dispute information into clearer analytical outputs, supporting risk identification and corrective action planning. Its value lies in making patterns and risks easier to detect and communicate in a banking context.