Security and data
There's no single right architecture. There's yours, your constraints and your level of sensitivity. What follows is what we never compromise on, and what we decide with you.
It's used to run your project, and nothing else.
Your documents, your data, and the code produced for you.
We never open broader access than the scope requires, and we list all granted access from scoping.
Someone who can't access a document today can't access it through the AI either.
Every answer links to its source, and every operation leaves a trail you can audit.
Data processing is defined during scoping, and we provide whatever your DPO or legal team requires.
Three questions we address during scoping, project by project.
On our infrastructure, on your cloud, or entirely on your premises. Hosting on your own infrastructure is something we know how to do: it's what we already do for some of our own tools. What it involves depends on your hardware and the chosen model, so it's assessed during scoping rather than on a web page.
We're not tied to any vendor. Depending on the use case, the constraints and the level of confidentiality, we use a commercial model or an open model installed on your premises. The model is a component: we can switch later without rebuilding everything, so you're not at the mercy of a vendor's pricing policy.
Everything, some of it, or nothing. It's a slider, not a checkbox, and you're the one who sets it.
Because what suits a spare parts manufacturer doesn't suit a defense subcontractor. A vendor who announces its architecture before knowing your constraints is selling you its product, not your solution.
We work with yours, and we write it into the scoping before we start.

It will. That's its job.
We'd rather address them before we start than at go-live. Bring them into the loop from scoping: we answer point by point and document what's agreed.
Our job is to keep you one step ahead.
Thirty minutes to understand your situation. At the end, we tell you whether we can help, and if so, how.
We support manufacturers in France, Europe, the Middle East and Africa.
Aerospace
Manufacturing
Plastics processing
Material handlingIt depends on where you start, but you always get a price and a date in writing before we begin. For reference: the AI audit is €3,000 excl. VAT, and half of it is credited toward your first project with us. The Bpifrance AI Diagnostic is 40% funded. Training a team in a Bootcamp is €2,500 excl. VAT per group, fundable by your OPCO. A ready-made agent is a setup fee followed by a monthly fee. And custom development is a fixed fee, with no subscription, priced after a few days of scoping.
For an agent already built at other manufacturers, allow 2 to 6 weeks before it runs at your site. For custom development, we scope for a few days, then move forward in two-week sprints: at the end of each sprint, you see something concrete. A heavier project, such as production scheduling, usually takes two to three months.
Not necessarily. If you already know which workflow is costing you (your quotes, your service reports, your orders), we can go straight to an agent. If you do not know where to start, the audit is the right entry point: we come on site, with management and business teams, and identify where AI really pays off, and in what order. And the audit may well conclude that you should train your teams first, or use a tool other than ours.
Yes. That is the case for almost every manufacturer we work with: data in the ERP, in ten Excel files, in emails. Our agents were born in those conditions. We often start with simple exports from your tools to test the approach, then automate data collection once it works. If something is a real blocker, we tell you during scoping, before you commit to anything.
Your data stays yours. We decide with you where it is hosted from the scoping stage: on servers in Europe, or directly on your own servers if your business requires it. We can also install your own AI model on premises, so nothing leaves your company.
By putting them in the loop from the start. The people who will use the agent take part in scoping and test the first versions: that is how a solution gets used, not just delivered. We also train your teams, and we measure usage 30 days after each training session. Then we stay by your side, with a new use case every quarter.