AIthatworkswithoutsendingyourdataanywhere
We install an AI model on a computer in your office and train your team to get real work out of it. The document you paste never leaves the building.
Why a school is telling you this, and not an IT firm
Because the hard part is not installing it. It is getting the team to use it well.
Setting up a local model is an afternoon of work. What decides whether it is worth anything is whether people know what to ask it, what not to believe, and when the answer needs checking. That part is training, and training is what we do.
How we do it
Four steps, and the first one is telling you if it is not for you.
We check whether it suits you
If you are a small team using AI occasionally, the cloud is cheaper. We will say so before you spend anything.
We pick the model and the machine
Based on what you want to do and how many people use it at once. There is no single right answer.
We leave it running
Model, a web interface for the whole team, and search over your own internal documents.
We train the team
The part that makes it actually get used. Eligible for FUNDAE funding.
Local, cloud, or both
There is no option that is right for everyone. This table is the conversation we would have with you.
| Cloud AI | Local AI | Both | |
|---|---|---|---|
| Where your data goes | To the vendor's servers | It stays in your office | Sensitive work local, the rest in the cloud |
| Cost | Per seat, per month, forever | Up-front hardware and electricity | Depends on real usage, and we work it out with you |
| Quality on hard tasks | Still ahead | More than enough for office work | Each one where it performs |
The AI Act does not go away just because the model is yours
Running AI in house saves you international data transfers and the contracts with the model vendor. It does not save you from informing people, from supervising the output, or from the Article 4 duty to train your staff, in force since February 2025. In fact, with no vendor in the picture, that part becomes entirely yours.
Who you will work with
Julien Bellet
Founder of IAescola. AI consultant and team trainer.
Julien Bellet works with companies and management teams to bring artificial intelligence into their processes, in Catalan, Spanish, English and French. Based in the Penedès, serving all of Catalonia.
LinkedIn profileFrequently asked questions
What does local AI actually mean?
The model runs on a computer inside your company. When you paste in a contract, a payslip or a client email, that text never leaves the office. No external service receives it and none can store it.
Does it work in Catalan?
Honestly: large open models understand and write Catalan, but there is currently no public, systematic evaluation measuring that quality, unlike Spanish. That is why we test it on your own documents before recommending anything. The Spanish government and the Barcelona Supercomputing Center have announced a model trained on Spanish and the co-official languages, but there is still no availability date or published results.
Is it better than ChatGPT or Copilot?
For everyday office work, summarising, drafting, translating and answering questions about internal documents, it is more than enough. For complex reasoning or very long contexts, paid cloud models are still ahead. Anyone telling you otherwise is selling something.
When is it NOT worth it?
If you are five people using AI now and then, the investment does not pay for itself and the cloud is cheaper. It is the first thing we check, and if that is your case we will tell you.
What does it cost?
It depends on the hardware you need and the size of your team. We do not publish prices because every case is different, but you get a fixed quote before anything starts.
Who maintains it afterwards?
Models and system need updating over time. We leave you the documentation and train someone in house to run it, or we run it for you. That is decided up front, not later.
Does this remove legal obligations?
Some yes, many no. International data transfers and the processor contract with the model vendor go away. What stays: the GDPR principles, impact assessments where required, records of processing, telling people when they are talking to an AI, and the duty to train your team. Local does not mean exempt.
Can I use FUNDAE funding as a self-employed worker?
If you are self-employed with NO employees, no: the funding comes from the vocational training contribution paid under the general social security scheme, and the self-employed scheme does not pay it. If you have employees, you can fund their training. Self-employed workers on their own have subsidised training calls instead, which is a different route, and we offer direct payment with a group discount.
How would you like to start?
We adapt the entry point to your situation.