Answers from your own documents, with every claim cited to its page
A RAG system (retrieval-augmented generation) built to your scope, for a business whose documents have outgrown search. Every source, your permissions, and a pass mark agreed before launch.
No list price. It starts with a diagnostic at a fixed fee, and the build is quoted in writing after it.
Seven days before the date, in full. After that, a credit.
When you need a knowledge system
And when you do not.
Retrieval-augmented generation (RAG) built to your scope: many sources, your permissions, and a test set of real questions. Knowledge lookup is the standard version. This begins where it stops.
You need it when
- More sources than one folder: drives, mail, a wiki, a case system, scans.
- Permissions that differ by team, client or matter.
- Questions that need several documents read together, or a figure found across them.
- A body of documents that changes daily and has to stay current.
You do not when
- One approved folder, a few hundred documents, a dozen staff: Knowledge lookup covers it at a list price.
- A question a search box already answers.
- Documents your own rules say may not be pooled.
What a knowledge system includes
Part by part.
- 01
Sources are mapped
Every store the system may read, who may see what in it, and how often it changes.
- 02
Documents are prepared
Scans read, tables kept as tables, each passage tied to its page and its version.
- 03
Retrieval is tuned
Keyword and meaning search together, then reranked, so the right passages come first on your questions.
- 04
Permissions are enforced
The index carries your access rules. A person retrieves only what they could open themselves.
- 05
Answers are cited
Each claim links to its passage. Where the sources are silent, the system says so.
- 06
A question set is written
Real questions from your staff with the right answers, scored before launch and after every change.
- 07
It is wired in
Chat, the intranet, the tools your team already works in, and an interface for anything else.
Works in tools such as
- Google Drive
- SharePoint
- Notion
- Slack
- Microsoft Teams
These are examples. We confirm the method for each of your tools before we build.
How a knowledge system is proved
A locked test set and a pass mark, agreed before the build.
- Answer and citation accuracy on your question set, against the pass mark agreed before the build.
- Permission tests on every release: no passage reaches someone who could not open the original.
- Coverage: the share of questions the sources could answer, and the gaps that need a document written.
How a knowledge system is run
After launch: monitoring, updates and what you own.
- Monitoring on every query: errors, slow answers, permission denials and unanswered questions raise an alert.
- Sources re-index as they change; releases are re-scored on the question set first.
- Handover: the pipeline, the index configuration, a runbook and a data map, in your name.
What a knowledge system will not do
Limits we set on purpose.
- It does not answer from the open internet or from anything outside the approved sources.
- It does not guess when the sources are silent. It says so, and can route the question to a person.
- It does not pool documents your own rules keep apart.
- It is not shown to customers unless you decide it should be.
Questions about a knowledge system
How is this different from Knowledge lookup?
Knowledge lookup is the standard system: one approved source set, a few hundred documents, a dozen staff, the tools its page lists, at a list price. A knowledge system is built to your scope when one of those does not hold: more sources, permissions that differ by team, documents that change daily, or questions that need several documents read together.
Which model does it use?
The one that fits your rules. The model is chosen with you, inside the data commitments you have made, and can run inside your own account through private inference.
Can it find a figure across our documents?
Where the figure is in the documents, yes, and it shows the passages it used. It does not estimate what the documents do not hold.
What do you need from us?
Access to the sources, a map of who may see what, and a set of real questions with the right answers from the people who know. The question set is the proof the system is judged against.
Talk to us about knowledge systems
Fifteen minutes on a call is enough to tell whether it fits your work. If it does not, we say so.