AI in your company, without your data leaving it
TIC-NOVA installs open-source AI models on your own infrastructure. Your teams get an assistant, search across your documents and a coding aid. Your data stays with you.
Proof of concept from €8,000.
AI and confidential data: the roadblock for sensitive sectors
Your teams want to use AI. Yet your patient records, your contracts, your customer data or your source code must not leave for a cloud service you do not control.
Healthcare, finance, industry, the public sector, law or consulting firms: the risk is not theoretical. As soon as an employee pastes a document into an online tool, the information leaves your perimeter. You no longer know where it is processed, or who can access it.
Two answers exist. Ban AI, and your teams work around the ban. Or bring it closer to your data. That is what we do.
Sovereign AI in the enterprise: AI installed on your side
In 30 seconds: we install AI on your machines, like any other piece of software. You use it like an online assistant, but what you entrust to it does not leave your company.
The version for a CTO
We deploy an open-weight language model, such as Llama or Mistral, on your infrastructure. An inference server exposes it to your applications through an internal API. To query your documents, we add semantic search (RAG): your documents are split and indexed in a local vector database, then the relevant passages are provided to the model at the time of the question.
The model weights, the index, the logs and the questions stay on your network. The model does not learn from your exchanges: there is no retraining behind your back. Authentication can rely on your corporate directory, so that access rights to documents follow your own.
Three concrete uses: assistant, document search, coding aid
-
Internal AI assistant
Your teams ask questions in plain language: procedures, standard contracts, internal memos, meeting minutes. The assistant answers from your documents and shows where the answer comes from.
-
AI document search
You search by meaning, not just by keyword, across your documents, your files and your knowledge base. Every result points back to the original document, which you can check.
-
Local coding aid
Your developers get a coding assistant that runs locally: completion, explanation, tests, review. Your source code does not leave for a vendor.
For code, we deploy coding assistants locally, connected to a model hosted on your side. The principle is the same as the assistants on the market, without sending your code outside.
A proof of concept from €8,000, before any commitment
We start small, on one use case, with your documents. You judge on the evidence before going further.
-
Scoping
Together we choose a use case, the documents involved and the success criteria.
-
Deployment on your side
We install the model and its tools on your infrastructure: your servers or your private cloud.
-
Evaluation
Your users test on your real documents. We measure quality with them, then we adjust.
-
Handover
We give you the code, the documentation and the access, so your teams can run the solution.
We work in sprints of 1 or 2 weeks, with a usable version and a review with you at the end of each sprint. See our method.
What “sovereign” concretely means for your data
The word is often vague. Here is what it covers in our projects, point by point.
-
Where the models run
On the infrastructure you designate: your servers, or a private cloud you control.
-
Where the data stays
Your documents, your questions and the answers stay within that perimeter. Nothing is sent to an external AI API.
-
What leaves the perimeter
None of your data. Only model and update downloads come in, and we set those up with you.
-
Who stays in control
You do. Access and documentation are handed over to you, and the code remains yours.
On-premise LLM or private cloud: your choice, and room to grow
You can host the model on your servers, or in a private cloud you control. In both cases, the infrastructure is yours and the data stays within your perimeter.
We design the solution to evolve: we start with one use case and a small number of users, then add machines, models or uses as the need grows. We scope the hardware with you at the start, not after the invoice.
What we do not promise
A local AI has real advantages. It also has limits. Better to know them beforehand.
- An open model hosted on your side can perform worse than the largest cloud models on some tasks. We measure it on your use cases during the proof of concept, not on promises.
- Hardware has a cost. A model of useful size needs graphics cards or a suitable server. We estimate that need with you before any commitment.
- Deploying on your side does not, on its own, make your organization compliant with a regulation. It reduces data flows to the outside. Compliance remains work to do with your security officer and your legal counsel.
Looking instead to strengthen or equip your development teams? See our offers for companies.
Frequently asked questions about sovereign AI
Where is my data hosted with a sovereign AI?
With you. The model, its tools and your documents are installed on the infrastructure you choose: your servers or your private cloud. We configure the solution without calls to an external AI service, and the access is handed over to you.
Which AI models do you use?
Open-weight models, such as Llama and Mistral, chosen according to your use case, your working language and your hardware. For code, we deploy coding assistants locally. The choice of model is confirmed during the proof of concept, on your documents.
How does the quality compare with a cloud AI?
On some tasks, the largest cloud models remain more capable. On others, such as answering from your documents, a well-integrated open model can be enough. We do not claim it in advance: the proof of concept measures it with your users.
What hardware do I need to run an LLM in-house?
It depends on the size of the model and the number of users. A model of good size usually needs a server equipped with graphics cards. We size the need at scoping, and we can deploy on existing hardware or on a private cloud.
What does it cost beyond the proof of concept?
The proof of concept starts at €8,000. Beyond that, the cost depends on the scope, the hardware and the number of users. We price what comes next with you after the proof of concept.
Let’s talk about your use case
Tell us about your documents, your constraints and your infrastructure. We will tell you whether AI hosted on your side can meet your need, and at what cost. Write to TIC-NOVA.