Glossary
A technique where a language model's response is grounded by first retrieving relevant information from an external knowledge source — typically a vector database — then using that retrieved context to generate a more accurate, current answer.
Why It Matters
A language model's built-in knowledge is frozen at training time and doesn't know a business's private or current data. RAG lets a model answer questions using a company's actual, up-to-date documents and data, instead of relying only on what it learned during training — meaningfully reducing fabricated or outdated answers.
This is part of our AI Systems capability — the infrastructure that lets intelligence get embedded into a workflow, not bolted on as a standalone feature.
See the full AI Systems capabilityRelated Terms
Does RAG eliminate hallucination completely?
No — it significantly reduces it by grounding answers in retrieved facts, but doesn't eliminate the risk entirely. Good RAG implementations still need to handle cases where the retrieval step returns weak or irrelevant results.
How is this different from fine-tuning a model on custom data?
Fine-tuning changes the model's underlying weights and is expensive to update; RAG retrieves fresh, current information at the moment of the query, which is faster and cheaper to keep up to date.