Model Classification

Purpose

Model Classification is designed to identify how a model has evolved between two dates by reviewing its documentation package, reconstructing the criteria that supported its previous classification, and proposing either a new classification or confirmation that no change is needed.

Primary users

The primary user is specified as Both. No further user segmentation is provided, but the agent is associated with AQS and supports users involved in model classification review, governance validation, and documentation of assessment work.

Where it fits (process/stage/trigger)

The agent fits into a model governance or classification review process when two model documentation packages, one at T0 and one at T1, must be compared and assessed against the previous classification record and applicable AQS governance or classification policy rules.

Key capabilities / workflow

The workflow reviews the model documentation packages, identifies model evolution between the two dates, reconstructs the prior classification rationale and criteria, applies the relevant governance or classification policy rules, proposes a new classification or confirms no change, documents the work performed, and routes the result through a HITL UI for validation.

Inputs

The inputs are the Model Document package T0, the Model Document package T1, the previous classification record, the historical rationale or criteria used for the prior classification, and the governance or classification policy rules applicable in AQS.

Outputs / Deliverables

The outputs are a documented assessment of model evolution, a reconstruction of the prior classification criteria and why they applied, a proposed new classification or confirmation of no change, and a complete work log or audit trail suitable for governance and review, validated through a HITL UI.

Value

The value of Model Classification is to support a traceable and documented classification review by connecting model evolution evidence, prior rationale, applicable policy rules, proposed classification outcomes, and human validation in a governance-ready audit trail.