RAG
RAG, short for retrieval-augmented generation, has an AI look things up in your own documents first and only then give an answer. That way the model works from your facts instead of from whatever it learned at some point.
A language model doesn't know your business on its own, and sometimes it makes up something that sounds plausible. RAG fixes that by first giving the model the relevant pieces from your knowledge base, and the model then bases its answer on those. It reasons about your current facts instead of an outdated memory.
This is the standard way to make an AI assistant reliable about a specific business, without having to retrain the whole model.
In practice.
The AI receptionist at Inge de Bruijn Zwembad works this way. It first looks up the answer in the pool's knowledge base and only then writes a reply, so it doesn't guess.
Frequently asked questions.
What does RAG mean?
RAG stands for retrieval-augmented generation. The AI first looks up the relevant pieces in your own documents and bases its answer on them, instead of drawing on whatever the model learned at some point.
Why does RAG matter for a business AI?
Because it lets a model answer based on your current facts instead of a general, possibly outdated memory. That makes the assistant more reliable and lowers the risk of made-up answers.
Related terms.
Want to see how this works in practice? Have a look at AI assistant and AI receptionist.
Curious what this could do for you?
Walk me through one process. In half an hour I will show you where this kind of work gets stuck in your business, and what it would save to hand it off.