Model quality assurance P&C - Model Stability Over Time (Drift Detection)

Purpose

Model quality assurance P&C - Model Stability Over Time (Drift Detection) supports model quality assurance by scoring AI models on quantitative stability and bias metrics, then using GenAI to generate qualitative commentary based on the available validation information.

Primary users

The primary user is not specified in the provided information. The agent is associated with the AQS team/BU and is owned by Ronan Davit.

Where it fits (process/stage/trigger)

This agent fits within insurance model quality assurance activities, particularly when model metadata, validation KPIs, and drift or bias metrics are available for review over time.

Key capabilities / workflow

The agent analyzes model metadata, validation KPIs, and drift or bias metrics, scores models on quantitative stability and bias indicators, checks whether the available metrics are sufficient, and generates qualitative commentary for quality assurance review.

Inputs

The specified dataset includes model metadata, validation KPIs, and drift/bias metrics. Other inputs are not specified.

Outputs / Deliverables

The provided use case indicates that the agent produces AI-generated scores on quantitative stability and bias metrics, as well as GenAI-generated qualitative commentary. Other outputs are not specified.

Value

The agent helps support insurance model quality assurance by combining quantitative assessment of model stability and bias with qualitative commentary, enabling more structured review of model behavior over time.