A private assistant onyour own knowledge.
We build assistants that answer from your documents and data, running in your cloud or your data centre, with nothing sent anywhere you have not approved.
01Who it is for
The problems we are usually called about.
Organisations with sensitive documents and strict data rules, whose staff still need the help that modern AI assistants give.
Knowledge locked in documents
Policies, contracts and procedures run to thousands of pages, and finding the right paragraph takes an expert.
Public AI tools are off limits
Security or regulation rules them out, so staff either go without or use them quietly.
Answers without evidence
An assistant that cannot show where an answer came from is no use in a regulated setting.
Everyone sees everything
Simple tools ignore who is allowed to read which document.
02What we do
Your knowledge, your infrastructure.
Answers with sources
Retrieval over your policies, contracts and procedures, citing where each answer came from.
Private deployment
Models hosted in your own cloud account or on your own hardware.
Access control
People see answers only from documents they are allowed to read.
Accuracy evaluation
Tested against questions your experts have already answered.
03How it runs
Step by step.
- 01
Scoping call
We agree which documents and which users come first, and where the system may run.
- 02
Deployment choice
We decide between a model in your cloud account and one on your own hardware, based on your rules.
- 03
Index your content
Documents are processed and indexed inside your environment, with their access permissions.
- 04
Evaluate
Your experts supply real questions and answers, and we measure the assistant against them.
- 05
Launch and maintain
Staff start using it, with feedback captured and content kept up to date automatically.
04What you get
Delivered, documented, yours.
An assistant your staff can trust
Because every answer shows its source.
A system you own
Running on your infrastructure under your policies.
An accuracy report
And a way to keep measuring it.
Tools we work with
Your stack, not ours
From our work
Account reconciliation platform
A month-end reconciliation platform for a finance team moving off spreadsheets. AI proposes the transaction matches; people review and sign off.
Questions
What clients ask first.
Does any data leave our environment?
That depends on the deployment you choose. With a model hosted on your own infrastructure, nothing leaves. With a model in your cloud account, data stays within that account under your agreement.
How do we know an answer is right?
Every answer cites the passages it used, so a reader can check it in seconds. We also measure accuracy against questions your experts have already answered.
Does it respect document permissions?
Yes. A user only receives answers drawn from documents they are allowed to open.
How does it stay up to date?
New and changed documents are re-indexed automatically, and removed ones are dropped.
Are private models as good as the public ones?
For many document tasks they are close enough, and the gap is narrowing. We test candidates on your own questions so the choice rests on evidence.
Tell us what you run.
Thirty minutes with an engineer. You leave with a clear scope, a timeline and a recommended approach.
