AI product & model operations · Core

Inference

Also called: model inference

ELI5

The trained model produces an answer for new data.

Used in conversation

“Inference latency must stay below two seconds.”

Definition

Use of a trained model to generate a prediction or output from new input.

Here “Inference” means: Use of a trained model to generate a prediction or output from new input.

Pitch context

You will see “Inference” in product strategy decks, roadmaps, experiments and architecture reviews when the discussion reaches ai product & model operations.

Why it matters: In ai product & model operations, the scope can change what data may be used, who may access it and which controls are required.

Sources & evidence · 2

Direct term-level sources and supporting source families.

National Institute of Standards and Technology — Computer Security Resource Center GlossaryDirect source · primary · checked 2026-08-16
Scrum Guides — The Scrum Guide — official current versionSupporting source family · primary · checked 2026-08-16