Private AI
Ask your own business. Keep the answer in the building.
A local AI model on hardware you own, answering from approved company documents with the source shown. Files, questions and answers stay inside your network.
Local processingThe model does not need the internet to answer
Source-linked answersPeople can inspect the document behind a claim
Permission-awareUsers only search material they are allowed to see
On-premise AI, local LLM, self-hosted AI
Different phrases. The same practical arrangement.
The model runs on a machine you control. Your approved files are indexed there. Questions and answers remain on your network instead of travelling to a public AI service.
Your documentsFiles, scans and spreadsheets
→
Your CortexSearch, model and permissions
→
Your teamAnswer with a visible source
No external processing step
What businesses use it for
The questions currently answered by searching or interrupting somebody.
Private AI is most useful when the answer already exists inside documents the business owns.
The demonstration
Two questions.
Ask one question you already know the answer to. Inspect the cited source. Then disconnect the internet and ask another.
Cortex
A private appliance, not another account.
Cortex combines a GPU machine with the software that indexes approved shares, searches them and produces a source-linked answer. It is installed on your network and bought outright.
Hardware and local model
Document indexing
User permission mapping
Installation and team training
Cortex starts from £4,950 installed. The first three months of Care are included; ongoing Care is optional.
The honest boundary
Private does not mean automatically perfect or secure.
A badly configured server is still badly configured, and an AI answer still needs checking. Owning the system changes who controls and can inspect the boundary; it does not remove the need for one.
Keep professional judgement human
Source links make verification practical. They do not transfer accountability to the model.
Keep access rules explicit
Private AI should not let everyone search every document merely because the files share a network.
Use cloud AI when it is the better buy
For a small team, light use and non-sensitive work, a subscription may be simpler and cheaper. See the three-year comparison.
Questions people ask
Private AI, in plain English.
What is private AI?
AI running on hardware you control, using models installed there, so the documents you ask about do not need to leave your network.
Is private AI the same as on-premise AI?
Near enough for this service. On-premise specifies that the machine is physically in your building; private AI is the broader idea.
What is a local LLM?
A large language model running on your computer rather than a provider’s server. Modern open-weight models are capable of document search, drafting and summarising for many business tasks.
Does it work if the internet goes down?
Yes. Everything required to search and answer is in the building. Disconnecting the line is part of the demonstration.
Is self-hosted AI automatically more secure?
No. Configuration, access control, updates and backups still matter. What changes is that the boundary is yours to inspect and test.
What does it cost?
Cortex starts at £4,950 installed, bought outright. Optional Care is £195 a month after the included period. See full pricing.
A useful demonstration
Bring a question your existing documents can answer.
We will show the source, disconnect the internet and ask another. That is more useful than a slide deck.