On premises

Your AI installed on your premises.

A server installed at your premises — for example, an NVIDIA DGX Spark or a custom-built machine suited to your needs — lets your AI work on site. With the on-premises installation option, your documents, communications and data are processed on your equipment; any external connections are agreed with you. This is our main difference: we install it, and you stay in control. On-premises installation remains your choice: we can also host your AI or install it on a server rented for your project. The scope and arrangements are agreed with you on a case-by-case basis.

How it works

A machine on your premises, set up by us.

We choose the machine with you, we install it in your offices, we put your AIs on it and configure it for your tools. You are not buying yet another subscription: you own a tool.

A machine to your measure

Depending on the work you hand over to your AI and the number of people using it, the machine can be a compact ready-made solution — such as the NVIDIA DGX Spark — or a custom-built machine assembled for your needs. We do not sell hardware at a price beyond what your uses require.

Installed within your walls

The server is set up on your premises, in an office, a small technical room or a cabinet, depending on the space available. We handle the set-up, the settings and the connection to your tools.

Your AIs work there

The agreed roles — preparing your files, following up your customers, screening applications, checking information — are installed on the machine and connected to your documents. Your AIs answer even when your internet connection is down, within the limits of what has been installed.

Documented and handed over

A written file describes the machine, its installation, its backups and the everyday gestures. Someone on your team can be trained to take it over, or we can handle it according to the agreed arrangement.

Prepare the teams to take over

A choice that matches your resources and your ambitions

When on-site installation is called for, when hosting is enough.

Being on your own premises is not an obligation, it is a decision. We help you take it by looking at your data, your constraints and your resources — and then we do what you decide.

On-site installation is called for when…

…your data must not leave your walls: trade secrets, confidential files, sensitive data, the constraints set by your customers or your sector. It is also called for when you want to keep full control: keep working even without internet, depend on no foreign subscription, know exactly where your documents go. This is the choice of firms and manufacturers that cannot afford to wonder.

Hosting is enough when…

…you are starting out, your data is not very sensitive and you do not yet have a team to look after a machine. Hosting asks for less investment upfront and is in place faster; the AI is built there for the same precise job, with the same requirements. It is an honest way in: it suits many projects, and we tell you when that is the case.

And you can change your mind

Start hosted, then install on your premises once the project has proven itself: that is a path we know how to prepare. What matters is that the choice is made on an informed basis, in line with your resources today and your ambitions tomorrow.

Your data

Your data stays on your premises.

That is the heart of the matter. Documents, customer records, application files, production data: all of it is processed on the machine inside your walls. Nothing is sent elsewhere by default.

We put things plainly: some features may need an outward connection — for instance to look up information online. When that is the case, we flag it in advance, we describe what leaves and you decide whether to accept it. No quiet sending, no surprises.

We agree with you on the authorised documents, the people who can access them and the retention rules. Compliance also depends on your own practices; it does not follow from the choice of a machine alone.

How far the AI goes, and where it stops

After go-live

Plan for who keeps the machine running.

A machine installed on your premises needs some attention: backups, updates, the occasional failure. We plan for an owner, simple gestures written down plainly, and a return to a known-good version if something goes wrong.

We define with you what is monitored, what is backed up and how often backups are tested. A backup that is never tested is not a backup.

If you entrust that monitoring to us, the proposal states precisely what is included and how we intervene. No response time and no service level is assumed: they are discussed, then written down.

Practical questions

Space, budget, change.

How much space is needed?

Less than you might think. Depending on your needs, the machine can fit in an office or require a small technical room. We choose the size according to the work you hand over to your AI, not the other way round, and we tell you clearly what space and what power supply are needed before the installation.

How much does it cost?

We do not publish prices by the kilo: each installation is priced after we have seen your uses and your premises. What we can say: on-site installation requires an upfront hardware investment, in exchange for which your data stays on your premises and you escape subscriptions that pile up. It is a choice matching your resources and your ambitions, and we present it without pushing you either way.

What happens if the machine breaks down?

Backups make it possible to restart on a new machine. Depending on the maintenance agreement, we intervene for the replacement and the restore, or we train your team to do it. The failure scenario is planned before go-live, not at the moment it happens.

Do we need an in-house technical person?

No, if you entrust the monitoring to us. If you prefer to keep everything in-house, we train a referent and hand over the documentation. Either way, someone has to be named: a machine without an owner ends up an abandoned machine.

Let us talk about what you want to keep on your premises.

Tell us what work you would like to hand over to your AI, which data is concerned and what your constraints are. You do not need to have chosen the machine: that is our job.