Answers, not a list of files
Asked in plain language, answered in plain language, from your own policies and contracts.
Two colleagues stop being the search engine
Service
A private system that answers questions from your own documents
Your policies, contracts, manuals and past work already contain the answers. We build the system that finds them and answers in plain language, with the source attached, without any of it leaving your control.
In short
A private knowledge system lets a team ask questions in plain language and get answers drawn from the organisation's own documents, with a link to the exact source. Unlike a public assistant it only reads material you have given it, and unlike a search tool it answers the question rather than returning a list of files to read. Because the documents are usually confidential, where the system runs and who can reach it matters as much as how well it answers.
Works with the systems you already run
Claude
SharePoint
Google Drive
Notion
iManage
AWSThe problem
The answer is in a document nobody can find
It is in a policy from three years ago, or a contract in a folder named after a project that got renamed, or a manual somebody saved to their own drive. The organisation has the knowledge. It just cannot be reached in the time available, so the same questions get asked of the same two experienced people until those people become a bottleneck.
Search does not solve this because search returns documents. What people want is the answer, plus enough of the source to check that it is right.
It breaks in four places, and the people doing it feel every one.
- 01Two people are the search engine
The same colleagues get asked everything
They know where things are and what the policy actually means. That makes them essential in a way nobody planned, and it means their week is interrupted constantly.
- 02Answers without sources
A confident answer nobody can verify is worthless here
In regulated work an answer you cannot trace is not usable. If the system cannot show which document and which page, the answer cannot be relied on for anything that matters.
- 03Confidentiality rules it out
The documents cannot go to a public tool
Client files, contracts, personnel records. The obvious tools are off the table not because they answer badly but because of where the documents would have to go.
- 04Permission blindness
A search that ignores who is asking
A system that reads everything and answers everyone will eventually tell somebody something they were not supposed to see. Then it gets switched off, correctly.
What changes
The same week, run differently
How it runs nowHow it runs after
Two experienced people are asked everything.
Most questions are answered without interrupting them.
Search returns forty documents to read.
The answer comes back with the page it came from.
Confidential material rules out the obvious tools.
It runs where your documents already live.
Everyone sees everything, or nobody uses it.
People see only what their existing permissions allow.
What we build
What the system does with your documents
The source, every time
The document and the section it came from, so anyone can check it in seconds.
Usable in regulated work
Your permissions, respected
People get answers drawn only from material they could already open.
Not a way around access control
It admits what it does not know
Where the documents do not cover the question, it says so rather than assembling something plausible.
Silence beats invention
Proof
The build that matters here is the boring half: working out which documents are authoritative, how permissions map onto answers, and what the system should do when the files genuinely do not cover the question. The answering is largely solved. Whether the answers can be trusted and shown to an auditor is the part that takes the work.
How it works
From first call to running system
- 01
We work out which questions are actually being asked
Usually a smaller set than expected, and often concentrated in one or two areas. That set defines what the system needs to be good at.
- 02
We agree what it may read, and who may ask
Which folders, which document types, and how existing permissions map onto answers. This is a governance decision and it comes before the build.
- 03
We build it where your documents already are
Inside your environment where that is what confidentiality requires, reading from the stores you already use rather than needing a copy elsewhere.
- 04
We test it against questions with known answers
Real questions where your experts already know the correct response, so accuracy is measured rather than assumed.
Confidence & control
What happens when the system is unsure
- Every answer carries its source
- The document and the section it came from, so anyone can check it in seconds. An answer without a source is not usable in regulated work and we do not build systems that produce them.
- It respects the permissions you already have
- People get answers drawn only from material they could already open. The system does not become a way around your access controls, which is the failure that gets these projects shut down.
- It says when it does not know
- Where the documents do not contain the answer, it says so rather than assembling something plausible. Being told there is nothing on file is a useful answer; a confident invention is not.
- You own it when we leave
- It is built inside your own accounts and your own cloud. If you never speak to us again it keeps running, and another team could pick it up. You are not renting your own process back from us.
Scope
What an engagement covers
Included
- A review of the questions actually being asked and by whom
- An agreed scope of what the system may read and who may ask
- The system itself, reading from your existing document stores
- Answers with citations back to the source document and section
- Permission handling aligned to your current access rules
- An accuracy test against questions with known answers
Not included
- Training a model on your documents. The system reads them at question time, which is safer and easier to change.
- Reorganising your document storage, beyond what is needed to read it reliably.
- Answers on material you have not given it access to.
- Legal or compliance sign-off on the answers it produces.
Scope, timeline and price are agreed after a short call, never before. We write down what the system has to do, and what it costs, before any build starts.
Questions
Frequently asked
Is our data used to train a model?
No. The system reads your documents at the moment a question is asked and uses them to answer. Nothing is trained on them, which means removing a document removes it from the answers immediately.
What stops it inventing an answer?
It answers from retrieved passages and cites them, so an answer with no source is visible as such. Where the documents do not cover the question it is built to say so, and that behaviour is tested rather than assumed.
Can it respect our existing permissions?
Yes, and it should. Answers are drawn only from material the person asking could already open. A system that ignores this will eventually surface something it should not, and that is how these projects get shut down.
Where does it run?
Inside your own environment if confidentiality requires it, so documents never leave systems you control. Where the material is less sensitive there are simpler options, and we would talk through the trade-off rather than assume.
How do you know it is accurate?
We test it against real questions where your experts already know the right answer, and report where it got them wrong. Accuracy claimed without that test is just a claim.
What does it cost?
Every engagement is priced to its own scope, so there is no list price. After a short discovery call we agree in writing what the system has to do and what it costs, before any build starts.
Related
Other things we build
Bring us the workflow that keeps eating your team's week.
Let's find the first one to fix.
The audit is free. If we can't find automation worth more than it costs to build, you owe us nothing, and you keep the roadmap.
Prefer email? info@chronexa.io
Or tell us what's slow
We'll review your workflows and come back with where AI saves the most time and cost.