AskQuo
RAG agents grounded in a company's own knowledge
A platform for building AI assistants that answer from an organisation's own documents and systems, with the answer traceable back to the source it came from. Teams point it at what they already have rather than writing anything new.
The problem
Company knowledge sits in thousands of files nobody can search across. The same questions get asked and re-answered by whoever happens to know, and the cost lands on the few people who do.
What made it hard
A general model will answer a question about your business confidently and wrongly. The engineering problem is not generating text — it is refusing to generate text that the retrieved sources do not support, and making the boundary between the two visible to whoever is reading.
What we built
A retrieval layer over the organisation's documents, feeding a generation step that cites what it used. Permissions are enforced at retrieval rather than at display, so an assistant cannot summarise a document the person asking is not allowed to open.
Architecture
- Document ingestion across mixed formats into a vector index
- Retrieval-augmented generation, answering only from what was retrieved
- Citations that point back to the source passage, not just the file
- Role-based access enforced at retrieval, so permissions cannot leak through a summary
- Deployable against existing document stores rather than requiring a migration
The product
What became possible
- An assistant can be pointed at a document set and answer from it, rather than being trained on it and then guessing.
- Every answer carries its sources, so a reader can check the claim instead of trusting the model.
- Retrieval-time permissions mean the same assistant serves people with different access without a separate deployment per audience.