On-premise AI
A private AI server
that sits in your own office
The model, the records and the chat history sit on a computer you own. Nothing is sent to an outside AI service, and there is no monthly software subscription.
Most AI tools send your work somewhere else to be read. A private AI server does the same job on a machine in your own office, on your own network, under your own lock and key.
It is one computer, usually small enough to sit on a shelf. It runs an open model, reads the records your business already keeps, and answers questions from them. Your team talks to it in the apps they already use. Nothing travels to an outside AI service, and when the internet goes down it keeps working.
Your data stays in your building
The model, the chat history and the records all live on the machine you bought. Nobody outside your office can reach them, and no outside service keeps a copy.
No subscription per person
You buy the machine and the build once. Adding the eleventh or the fiftieth person to the system does not change what you pay each month.
It reads what you already have
Attendance, projects, calendars, invoices, drawings, site photos and the files on your drive. It answers from them and says which record the answer came from.
It works in your apps
WhatsApp, a team chat, email, a sheet. Your staff carry on sending the same messages they already send. Nobody is asked to learn a new screen.
It asks before it acts
Anything that changes a record or sends a message is confirmed by a person first, and written into a log afterwards.
It keeps running without you
It restarts itself after a power cut and tells me when it cannot reach the internet, which is how my own studio has run since I built it.
What actually sits in the room
A small server, chosen for the size of your team and the work you want it to do. For most offices of ten to a hundred people this is a quiet box on a shelf with no screen attached to it. For a larger company, or for work that needs a model trained on your own material, it is a bigger machine with a graphics card in it.
On that machine sits an open model that runs locally, a copy of the records it is allowed to read, and the Agents that do the work. I set it up, you own it. The hardware is bought in your name and stays in your office.
- A machine sized to the team, bought in your name
- An open model that runs on it, with no outside account
- Read access to the records you choose, and nothing else
- Agents inside WhatsApp, team chat, email or a sheet
- A log of every action the system took
Why owners ask for this
The usual reason is simple. Someone has been pasting contracts, salary sheets or client drawings into a chat window, and the owner has started to wonder where all of it went. Once a business keeps its own records inside its own walls, that question stops being asked.
India's data protection law, the Digital Personal Data Protection Act 2023, has made this a board-level question rather than an IT one. I am not a lawyer and this page is not legal advice. What I can tell you is exactly where every piece of data sits, which machine reads it, and who can reach it. That is usually what your auditor wants to see written down.
The second reason is money that leaks quietly. Per-person software costs grow every time you hire. A machine you own does not.
Where the AI actually earns its keep
I do not put AI into work that already runs well. A good system removes the writing, the chasing and the re-typing that nobody enjoys and nobody checks: the end-of-day report, the attendance record, the leave request, the follow-up message, the handover note, the file that should have been saved in the right folder.
The audit comes first, and part of what you get from it is a list of the things to leave exactly as they are. That list is usually longer than the build list, and it saves more money.
What decides the price
There is no rate card, because two offices of the same size can need completely different builds. The quote follows the audit, and it names the machine, the systems being built, what is connected to what, and the handover.
Four things move the number: how many systems you want connected, how much of your record keeping is on paper today, the size of machine the work needs, and whether a computer you already own can be used instead of buying one. Care afterwards, if you want it, is quoted separately and can be stopped.
How it runs
- 01
Audit
I sit with the people who do the work and map how it really moves. You get the map, the build list and the leave-it-alone list.
- 02
Design
We agree what gets built, in what order, and what the records should look like at the end. You sign that off before anything is built.
- 03
Build
I build it and run it with one team first, on your machine, until it survives a normal week without anyone babysitting it.
- 04
Handover
Training in person, a day on site while people use it, and then the machine, the code and the accounts are yours.
Built, not described
Questions people ask
Does a private AI server work without the internet?
Yes. The model runs on the machine in your office, so answers keep coming when the line goes down. Anything that has to reach the outside world, like sending a WhatsApp message, waits until the connection is back.
What hardware does an office AI server need?
It depends on how many people use it and what you ask of it. A small office running Agents and records can use a compact machine with no graphics card. Training a model on your own material needs a bigger one. The audit decides which, and you buy it in your name.
Is this cheaper than paying per user for AI tools?
It is a different shape. You pay once for the build and the machine instead of every month for every person, so the sum depends on your headcount and how long you keep it. I will show you both shapes honestly before you decide.
Can it read our WhatsApp groups and our files?
Only what you point it at. You choose which groups, folders and sheets it may read, and that list is visible and can be changed. It writes down what it looked at and what it did.
What happens if we stop working with you?
Nothing stops. The machine is yours, the code is yours and the records are yours. I hand over the accounts, the logins and the documentation, and your own IT person can run it from there.
Is an on-premise setup enough for the DPDP Act?
Keeping records inside your own building removes most of the hard questions about where data went, but the law covers more than storage. I document exactly what sits where and who can reach it, and your legal advisor uses that to judge compliance.
How long does it take to build?
The schedule comes with the quote, after the audit, because it depends on how much of your record keeping has to be cleaned up first. Most builds start with one team and one system rather than the whole company at once.
Next step