Contradictions
Two documents saying different things about the same thing.
Beta
Ragen Brain
Before AI can answer, someone has to put the knowledge in order
Ragen Brain does it from your documents, and a person approves the result
AFTER: KNOWLEDGE PAGES
Most companies don't have an AI problem. They have a scattered-information problem: documentation in three places, decisions buried in email, a procedure that exists in two versions and nobody knows which one is current.
How it works
Brain handles what happens before the question.
Feeding that into an assistant doesn't clean it up. It gives you faster access to the same scattered information.
-
Reads your documents
Brain reads the documents already in your knowledge base.
-
Extracts knowledge pages
Short write-ups of policies, processes, roles and products, where every statement quotes its source.
-
A person approves
Someone at your company assigns each page an owner and approves it.
-
Cited in answers
Published pages are cited in chat answers next to your documents.
EXAMPLE KNOWLEDGE PAGE
PROCESS Approved · published
Refund process
Owner: Head of Customer Service Open to: Customer service, Finance
The team lead confirms every refund before it is paid out.
“Each refund must be confirmed by the team lead before payment.”
customer-service-handbook.pdf · p. 12
The customer gets an answer within three working days.
“Customers receive an answer within three working days of the request.”
sla-2026.docx · §4
Refunds above the agreed limit go to Finance.
“Refunds above the agreed limit are passed to Finance for approval.”
email “Refund limits” · 12 March
- 1 · handbook.pdf
- 2 · sla-2026.docx
- 3 · email “Refund limits”
Read the full description
Ragen Brain is the part of Ragen that handles what happens before the question. It reads the documents in your knowledge base and extracts knowledge pages: short write-ups of policies, processes, roles and products, where every statement quotes its source. Someone at your company assigns each page an owner and approves it, and published pages are cited in chat answers next to your documents.
What nobody sees
Along the way it surfaces what usually nobody sees.
Gaps
A process with nobody named as the one who performs it.
Out of date
A source that was deleted, or whose new version no longer contains the quoted words.
The knowledge graph
Every page is a node and every relation a line.
- Thick line: backed by a quote from a source
- Thin amber line: uncertain
- Inferred by the model without a quote: hidden until you show them
Three rules it runs on
Brain proposes. Someone at your company approves.
Software that decides on its own what's true about your company isn't a feature, it's a risk.
Nothing is published without a person
No Brain page reaches the knowledge base until a person approves and publishes it.
Permissions set once, while you sort
respected on every answer afterwards
Not per file, but per knowledge page, together with its owner.
The result is yours, readable without us
Open the files in any editor and version them like code. If you stop at the cleanup, that's a result too.
The rules in full
No Brain page reaches the knowledge base until a person approves and publishes it. Brain proposes, someone at your company approves. Software that decides on its own what's true about your company isn't a feature, it's a risk.
You set permissions once, while you're sorting things out. Not per file, but per knowledge page, together with its owner. A new page is open to the people who could read its source, and Ragen respects that on every answer afterwards.
The result is yours, and readable without us. Approved pages download as a bundle of Markdown files with a graph of their relations, which you can open in any editor and version like code. Publishing them into Ragen's knowledge base takes one click, but you don't have to. If you stop at the cleanup, that's a result too.