Recalibrate probability-of-default models monthly using new portfolio performance data.

Recalibrate probability-of-default models monthly using new portfolio performance data. Detect drift in PD predictions versus realized defaults across segment, vintage, and macro factor. Propose adjusted PD weights when drift exceeds tolerance, with a written rationale the CRO reviews before sign-off. Past misses: we missed a hospitality-sector concentration shift through spring 2024, and held PD too low on a vintage with rising DPD in Q3.

Per-case agent output isn’t captured yet — recommendations and alerts will land here once the agent produces them.