Retrieval-augmented generation (RAG) means: find the relevant text first, then ask the model to answer using it. It is a good pattern. It is also easy to reach for when something simpler would do.
1. Does it all fit in the prompt?#
Modern models accept a lot of text. If your documents are a few pages, just include them. You will skip an entire system and probably get better answers.
2. Is the problem search or reasoning?#
If people mostly need to find the right page, a good search box may be the whole answer. Add the model when you need it to read, compare or summarise.
3. How will you know retrieval worked?#
Most RAG failures are retrieval failures: the right passage never reached the model. Log what was retrieved next to every answer, and check it by hand on a handful of questions.
Tip
Start with the plainest retrieval you can (keywords, or one embedding model) and measure before making it clever.
When it is worth it#
RAG earns its place when your content is large, changes often, and answers must point to a source. Build it then, and build the evaluation first.