The part of an AI system that does processing for the user. A peculiar name for sure; however, the inference term dates back to very early AI systems and it has not gone away, as yet.
AI inference refers to an AI model that does work for the user because it was previously trained on similar data. Also called "AI runtime," AI inference answers questions, generates content of all variety, makes forecasts, translates languages and more.
In a non-AI system, the inference counterpart is an application processing data; in other words, the reason for having computers. See inference engine and AI training vs. inference.