Internal Knowledge Assistant

The problem

At 30–50 people, institutional knowledge is scattered across a wiki, handbooks, runbooks, and policies, and half of it lives in a few people's heads. New hires and busy staff constantly ping colleagues with "how do we do X?", breaking focus and slowing onboarding. A plain chatbot would only make it worse, confident, unsourced, and often wrong.

The build

A grounded RAG assistant over the company's own internal documentation. Every document is chunked, embedded, and stored in a vector index with its source attached. A Gemini agent answers staff questions using only the retrieved passages, cites the exact source documents, and refuses when the answer isn't there, so answers are trustworthy and traceable. Conversation memory handles follow-ups, and it re-targets to any team by pointing at a new document folder.

Results

4,000+

documents indexed

80%

fewer internal questions

<3s

median answer time