B2B services
B2B services company
RAG assistants over corporate documentation
The client
A B2B services company with a body of technical and commercial documentation built up over years: the house know-how, spread across manuals, procedures and product documentation. By agreement with the client, we do not publish their name.
The challenge
That knowledge existed, but finding a specific answer required knowing in advance which document to look in, so the same questions kept coming back to the team. And there were two audiences with different needs: inside, people who need the detail of the know-how to do their work; outside, people arriving with a question about the company and its services. A generic chatbot served neither, because in this context a made-up answer is worse than no answer: it erodes trust in the whole system.
What we did
We built two RAG assistants, one for each audience. The internal one answers on the company's know-how, so the team finds in seconds what used to require digging through documents or asking a colleague. The external one serves website visitors with information about the company and its services. Both work the same way under the hood: they retrieve the relevant passages from the documentation and answer from them, always citing the source so whoever asked can verify it; and when the documentation does not cover the question, they say so instead of improvising. The deployment is private: the documentation never leaves their environment.
Technologies
Python · RAG · LLMs · vector databases · FastAPI
Result
The team consults the company's know-how without needing to know where it is written down, which also surfaces which parts of the documentation are worth improving. And website visitors get answers grounded in real information and traceable back to their source document, which turns the assistant into an entry point for new contacts.