AI Operations · Glossary
MLOps vs. AI Operations
MLOps governs the lifecycle of models. AI Operations governs the lifecycle of authored AI behaviour. The two disciplines are complementary, not adversarial, and sit at different layers of the stack.
MLOps emerged to govern the training, deployment, and monitoring of machine learning models. Datasets, model versions, drift detection, retraining pipelines. The substrate is the model.
AI Operations emerged to govern the authored behaviour that production AI executes. Policies, routines, evaluation suites, improvement loops. The substrate is the authored library.
A production AI system runs both. The model is governed by MLOps. The behaviour the system enacts, the policies it enforces, the routines it traverses, the tools it invokes, is governed by AI Operations.
Treating the two as one collapses authored behaviour into model behaviour. The result is a system that conflates model retraining with behavioural improvement. AI Operations keeps the two layers separate by design.