ML Models in Production: Why Enterprise AI Projects Stall

Six months in, fourteen ML models were live—but none had automated retraining. Three still used hardcoded paths from a laptop. When compliance flagging hit 40% false positives, an eleven-day investigation traced it back to an undocumented schema change. This is where enterprise AI projects actually break. Full article: https://davidohnstad.net/ml-models-production-enterprise-ai-stall/