Retrieval-augmented generation, usually shortened to RAG, is a design where an AI model is given relevant facts at the moment it answers. Instead of relying on memory from training, the system first searches your own content, then asks the model to answer using what it found.
The steps are simple. A customer asks a question, the system converts it to an embedding, searches a vector database or other index for the best-matching passages, and adds those passages to the prompt. The model writes a reply based on them and can cite the source.
RAG is the most common way retailers and venues build assistants that know their products, menus, returns policy, trading hours or staff procedures. Because the knowledge sits in your documents, you update it by changing the document, not by retraining a model.
RAG is not a guarantee of accuracy. If the search pulls the wrong passage, or your documents are outdated, contradictory or missing, the answer will be wrong. Good practice is to curate source content, show sources where possible, set the assistant to say "I don't know" when nothing relevant is found, and review a sample of conversations regularly.
RAG is usually cheaper, faster and easier to control than fine-tuning, which is why it is the default choice for business knowledge assistants.