RAG: the Open-book exam

There are three main approaches to making LLMs work with your data: Prompt Engineering, Fine-tuning, and RAG. Each has trade-offs, but RAG hits the sweet spot for most business applications.

Three approaches to custom LLM knowledge

Compare Prompt Engineering, Fine-tuning, and RAG.

Fine-tuning bakes knowledge into the model weights, which means every time your data changes you need to retrain. That is expensive, slow, and the model can still hallucinate because it "memorized" rather than "looked up" the answer. RAG keeps the data separate so updates are instant: just change the documents in your vector store.

Think of RAG as an Open-Book Exam. The LLM doesn't need to memorize every fact. Instead, when asked a question, it first looks up the relevant information in its "textbook" (your documents), then formulates an answer based on what it found. This is exactly how RAG works.

The four steps of RAG

Every RAG system follows this core flow.

Timed quiz: Quick check

Loading practice…

Ordering exercise: Order the RAG steps

Loading practice…

Matching exercise: Match approaches to descriptions

Loading practice…