The best explanation I can give as a non-expert is: it's used when you have a general-purpose LLM but want to give it some domain-specific knowledge. The query sent to the LLM is run through what's effectively a search engine that catches relevant terms etc, to find useful snippets of knowledge to send to the LLM alongside the query, so the query is _augmented_ with potentially useful information for answering the query.
The best explanation I can give as a non-expert is: it's used when you have a general-purpose LLM but want to give it some domain-specific knowledge. The query sent to the LLM is run through what's effectively a search engine that catches relevant terms etc, to find useful snippets of knowledge to send to the LLM alongside the query, so the query is _augmented_ with potentially useful information for answering the query.