Answers from your own files
Every business has knowledge trapped in documents: procedures, specifications, contracts, past quotes, the handbook nobody reads. A knowledge assistant makes that searchable by question rather than by filename.
This is the technique usually called retrieval-augmented generation, and the important thing about it is what it is not. It does not ask a public model what it thinks. It finds the relevant passage in your own material, answers from that, and shows you which document it used. Done properly, that makes a wrong answer visible rather than plausible.
Also sold as retrieval-augmented generation, RAG, enterprise search, AI knowledge base and document Q&A.
The work it takes
off your desk
Answer the question a new starter asks
The things everyone knows and nobody documented, answered from the material that does exist, without interrupting the one person who remembers.
Find the clause that matters
Across hundreds of contracts or specifications, with the source passage shown so somebody can check it.
Reuse what you have already written
Past quotes, method statements and tender responses, findable by what they said rather than what the file was called.
Support your own customers
The same assistant, pointed at the documentation you publish, answering from it rather than improvising.
No surprises,
in writing.
Scope and price agreed before anything is built. If we think you do not need this, we will tell you that instead.
Get a quote ↗- An assistant answering from your documents, with the source shown on every answer
- Permissions respected, so people only get answers from material they may see
- A defined refusal: it says it does not know rather than filling the gap
- A process for keeping the material current as documents change
- Your documents staying in your control, and never used to train public models
Where this
goes wrong
Every supplier will tell you what their service does. These are the ways it fails, and what we do about each one. If a supplier cannot tell you this, they have not built enough of them.
Confident wrong answers
The characteristic failure of this technology is fluent nonsense. Showing the source passage next to every answer turns an invisible error into an obvious one, which is why we do not ship without it.
Stale documents
An assistant answering from last year's procedure is worse than no assistant, because people trust it. There has to be a defined way material gets refreshed, and someone whose job that is.
Ignoring permissions
If the assistant can read the HR folder, it will answer questions from the HR folder. Access has to mirror the permissions people already have, and that is a design decision made at the start.
No graceful "I do not know"
A system with no refusal will always produce something. We set the threshold with you and tune it against real questions once it is in use.
What everyone
asks us
Yes. Retrieval-augmented generation is the technical name: retrieve the relevant passage from your own material, then answer from it. We use the plain description because it is what actually matters, which is that the answer comes from your documents and shows its source.
The usual business formats, including PDFs, Word documents, spreadsheets and pages from your own systems. Scanned material works too, with the same accuracy caveats as extraction.
Not if it is built properly. Access mirrors the permissions your people already have, and that is designed in at the start rather than bolted on.
No. Your material stays yours, and we put the specific arrangement in writing for the setup we propose.
Other AI services
Start with
one job.
Tell us the task that eats the most time. We will tell you honestly whether AI is the right answer for it, and what it would cost to find out.
Get a quotation