Playbook
We have 30 ML models in production and no idea which ones are drifting
A bank's ML team has deployed dozens of models with no shared lifecycle discipline. There is no central registry, retraining is ad-hoc, and drift detection lives in someone's personal notebook. The regulator has asked how the bank knows the models are still fit for purpose.
Trigger — Regulator review; lack of model lineage flagged.
Good outcome — Central registry, drift detection automated, retraining cadence governed, MLOps as a discipline.