Retrieval-augmented generation
Also called: RAG
ELI5
Find relevant sources first, then let the model answer using them.
Used in conversation
“Retrieval-augmented generation improves grounding but does not eliminate errors.”
Definition
A method that retrieves relevant external information and provides it to a generative model before output.
Here “Retrieval-augmented generation” means: A method that retrieves relevant external information and provides it to a generative model before output.
Pitch context
You will see “Retrieval-augmented generation” 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.
Need an investor or company deck?
Bring the business, the evidence or the half-formed ask. We help it travel beyond the spreadsheet.
See how pitch.dog can helpSources & 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