Model Risk Management demo
Every model on a control chart from production data, with a revalidation task the month it drifts. 1:14
What the demo shows
A model risk solution for banks: every model inventoried, monitored on live data, validated.
Seven agents score every model each month on production data against its own control limits and open a revalidation task, with its evidence, the moment one breaches; they also re-scope the inventory to SR 26-2, draft validation reports from the evidence and chase every finding — the Head of Model Risk Management and a validator decide and sign.
Transcript
- Model Risk Management watches and validates every model a bank runs for its model risk team.
- Each model gets a control chart, accuracy, stability, and overrides against its own limits.
- This morning, September's defaults arrive for the commercial real estate default model.
- Its accuracy line drops below the lower control limit, and a revalidation task opens on its own.
- The Drift point opens the evidence.
- Office loans defaulted at twice the predicted rate.
- Agents read 61 documents and draft the validation, with four findings and their proof.
- Under SR 26-2, the Rescope also cuts this year's validations from 214 to 131.
- Drift is caught the month it starts, not at next year's validation.
- Built on Botminds: agents, data, screens, and governance on one platform.
Transcribed from the film’s narration. Captions are on the player.