RAG – retrieval-augmented generation – is a method in which a language model first looks things up in selected documents and then formulates its answer on the basis of what it found.
A language model on its own only knows what it was trained on – and nothing about your organisation. With RAG it gains access to your own sources: the tone of voice guide, the annual report, the Q&As, the old press releases. It still answers in its own language, but with your content as the foundation.
That changes what the communications department can use AI for. Without RAG, you get generic text that could have been written for anyone. With RAG, you get answers built on what you have actually decided and written – and they are easier to check, because the source can be looked up.
In practice
A communications department can let an AI assistant answer internal questions based on the organisation's own policies and crisis-preparedness Q&As – instead of the internet's average. The answer comes with a reference to the document it is based on, so you can check it in seconds.
How to explain it to management
»RAG means the AI answers from our own documents instead of only from its memory – which makes the answers more precise and easier to check.«
If your organisation is about to connect AI to its own sources, that is a classic task for my AI advisory.