Skip to content

Glossary entry

Retrieval-Augmented Generation

A pattern in which a model uses retrieved external context at runtime to generate more grounded answers.

Why it matters

Retrieval-augmented generation matters economically because it can improve output relevance without always requiring fine-tuning, but it also adds context, data, and infrastructure cost that must be included in TCO.

RAG often shifts cost from model customization toward data pipelines, retrieval quality, and context assembly. That trade-off can be attractive, but only when the extra layers are measured properly.