Skip to main content

ML center

For: model risk, analytics, and senior risk.
Not for: daily approve/reject on individual applications.

In Decision PRO: ML center (role-gated).

What this is​

The ML center is where authorized teams review model performance, drift, champion/challenger status, and batch diagnostics. It supports governance, not front-line underwriting buttons.

Why it matters​

  • Detects when the world changed and models need attention.
  • Gives committees a shared place to discuss model health.
  • Separates monitoring from customer promises.

What you do​

  1. Pick the model family your bank tracks.
  2. Review dashboards against thresholds your governance committee defined.
  3. Open issues with the model owner when thresholds breach-attach dates and evidence your policy allows.

What you should not do​

  • Treat ML center tiles as customer-facing commitments.
  • Change production models from this page unless your bank’s change process explicitly allows it (usually it does not).

When to escalate​

  • Sustained drift or accuracy breaches.
  • Conflicting signals vs default prediction or portfolio delinquency.
  • Any proposal to use monitoring metrics as marketing claims.

Next: AI center · AIS explainability