Retrieval-augmented generation (RAG) retrieves selected information and supplies it to a generative model to help produce a response.
Definition
Retrieval-augmented generation combines a search or retrieval step with a generative model. The system finds material from selected sources, such as firm documents, and supplies it as context when generating a response. A firm implementing RAG may need to organize source files, manage permissions, configure retrieval, and review answers against cited material. Costs can include setup, access controls, document maintenance, and staff review, with the division depending on whether the tool is internal or provided as a service. RAG is not a guarantee that the model will find every relevant document or interpret retrieved material correctly. The generated answer can still be wrong, and a citation does not by itself prove that the answer follows the source. Source quality, document currency, access rules, and review procedures affect results. For project or client use, owners should define which repositories are allowed and who checks the response before relying on it.
How owners use it
- Limit retrieval to approved repositories and confirm permissions before indexing project records.
- Review generated answers against the retrieved source passages before using them.
- Example: 18 retrieved records checked at 5 minutes each require 18 x 5 = 90 minutes of review.
