RAG and AI agents with company data
Retrieval-augmented generation explained: how AI gets the right company knowledge, with permissions, through an API or MCP, and when RAG beats fine-tuning.
What RAG is
Retrieval-augmented generation lets a language model answer from your own documents: a search finds the relevant passages first, and the model writes its answer from them. The quality of the answer depends on the quality of that search.
RAG with Lookeen Server
Announced: Lookeen Server as a RAG layer makes email, files and Microsoft 365 content available to your own AI agents through an API or MCP, with each user's permissions applied. It is coming soon and not available yet. What Lookeen's AI does today is on the AI page.
Frequently asked questions
RAG or fine-tuning?
RAG keeps the knowledge in your documents and always answers from the current version; fine-tuning bakes knowledge into the model. For company data that changes daily and has access rights, RAG is usually the better fit.
Does RAG respect who may see what?
It has to: the search step must only return passages the asking user is allowed to read. That is why the retrieval layer matters as much as the model.
