External processing is removed from the answer path
Your question and source documents do not need to travel to a third-party AI service. Backups, support access and existing cloud storage still need their own decisions.
Why Local AI
Keep the AI close to the knowledge that makes the business valuable.
Not because the cloud is bad. Because when an AI needs years of prices, methods, drawings and customer history, keeping the processing in your building can remove risks and recurring costs instead of merely managing them.
Documents remain localNo external AI processor is needed to answer
One applianceNo per-seat licence for the team
Works offlineQuestions continue when the line is down
The problem nobody put on a list
Your team probably already wants an AI tool.
Somebody has likely pasted a quote, customer email, contract extract or drawing into a public tool because it was useful and no clear alternative existed. The issue is not bad intent. It is that commercially valuable material left the business’s control without a deliberate process around it.
If this is the immediate concern, the free 60-second policy builder gives staff a usable first boundary.
“You cannot stop people wanting useful AI. You can give them one the business controls.”
A policy handles today. A private tool can become the approved route tomorrow, where the use case justifies it.What changes when it runs locally
The boundary becomes something you can inspect.
Local processing does not make governance disappear. It makes the processing route, access controls and ownership easier to see and test directly.
Supplier terms cannot switch the appliance off
An installed open-weight model cannot be remotely deprecated or rate-limited. Updates are planned work rather than a silent dependency.
Team size stops multiplying licence cost
One appliance can serve approved users without adding a separate AI seat for every person.
The broadband line is no longer part of every answer
The model and indexed material are already inside the network, so document questions continue during an outage.
When cloud AI is the right answer
Often, honestly.
If a handful of people use AI lightly for general drafting and nothing especially sensitive goes into it, a cloud subscription is usually simpler. The case for ownership changes with the number of users, sensitivity of the documents and length of use — not with ideology.
What it looks like
One box.
Installed on your network, reading only the approved sources and answering without an internet dependency.
In practice
A source-linked answer your team can check.
People ask a plain-English question across the files they are permitted to search. The response points back to the source. Figures from spreadsheets are computed rather than improvised, and when the answer is not present the useful response is to say so.
Questions people ask
Local versus cloud, without slogans.
Is local AI better or simply more private?
For many SMEs the main benefits are control, privacy and predictable ownership. For general writing or research, a cloud tool may be the better experience and purchase.
What is the practical risk with public AI tools?
Staff may paste commercially sensitive work into a service without an agreed policy, record or deliberate decision about the provider’s terms.
Does the data really never leave?
Cortex does not need an external AI service to answer, and it works with the line disconnected. Existing cloud storage, remote support and backups remain separate routes to document and control.
How do updates work?
Updates are planned and tested rather than fetched silently by the appliance. That is part of the trade for keeping the answer path local.
Is owning cheaper than subscribing?
It depends on team size, duration and what people need to search. Our three-year comparison includes the cases where renting wins.
Decide from your use case
Bring one question and the documents that contain its answer.
We will tell you whether local AI, a cloud subscription or a simpler search tool is the sensible route.