"What's our refund policy for annual plans?" "Which form do I use for travel claims?" "Has anyone dealt with this customer before?" Every company has a few people who answer these questions all day, even though the answers are written down somewhere.
An internal knowledge assistant lets anyone ask in plain language and get an answer from your own documents in seconds, with a link to the exact page it came from.
How it works, without the jargon
- 1Your documents are indexed
- 2Someone asks in Slack or a web chat
- 3It finds the most relevant passages
- 4Writes an answer from only those
- 5Shows the source and section
The technical name is retrieval-augmented generation (RAG). All it means is that the AI looks things up in your documents first and answers from what it found, instead of from memory.
The three things that make it trustworthy
- Sources on every answer. If people can't check it, they won't rely on it.
- "I don't know." If the answer isn't in the documents, it says so and suggests who to ask. A confident wrong answer about policy does more damage than no answer.
- Permissions. If someone can't open the HR folder, the assistant shouldn't quote it to them.
Where it pays off most
- Onboarding: new hires ask the assistant instead of interrupting someone every ten minutes.
- Support teams: instant answers on policies and procedures during customer calls.
- Specialist knowledge: product details, compliance rules and past decisions that live in one person's head and a few old documents.
Keeping it current
It re-indexes automatically when documents change in Google Drive, Notion or SharePoint. Once a month, look at the questions it couldn't answer. That's your list of documentation gaps, and it's surprisingly useful on its own.
Questions people ask me
Is our data safe?
It can run entirely inside your own cloud account, and your documents aren't used to train public models.
What file types does it handle?
PDFs, Word and Google Docs, Notion pages, spreadsheets, and most wikis. Scanned documents need text extraction first, which we handle.
This is one of our custom AI builds. For larger teams, see enterprise.
Founder of Overtake. I was a founding engineer at four startups, two of them in Y Combinator. I write about the AI systems I build for small businesses: what works, what doesn't, and how to start.
About meFollow on X