Your team keeps asking the same questions, and the answers already sit in your documents. Searching for them is slow, and a general chatbot will answer confidently from nowhere.

What I build: An assistant that looks up the answer in your own files before it replies. It searches a folder you choose, such as Google Drive, PDFs or Notion pages. It answers from the passages it finds and names the file each answer came from. When nothing matches well enough, it tells the person asking and shows what it searched. Nothing is trained into a model: your files stay where they are, and the assistant reads them when a question comes in.

What $290 covers: One assistant on one collection of documents, answering in one place such as a web chat or Slack, with a nightly refresh that picks up new and changed files. More sources, user permissions or a customer-facing version are quoted separately.

What I need from you: Read access to the documents, 15 to 20 questions people really ask, and the answers you would accept.

How we check it: Before handover we run your questions through it together. For each one you see the answer, the cited file and how well it matched. Retrieval makes answers far more reliable, but it can still miss or misread a passage, so the source link is always there to check.

Handover: It runs on accounts you own. You get setup notes, how to add documents and how to read the question log, which shows what people ask and which documents are missing.

The cover shows my own Google Drive document assistant, built in n8n and walked through at fractionalhq.uk/use-cases/document-qa-citations. Each answer it sends carries the file it came from.