One AI for your whole company, and it stays inside your network.
Your staff draft, translate and work through documents the way they know from the well-known AI assistants. The model runs on hardware inside your network, no request goes to a model vendor, and the cost does not depend on how many requests people send. You run the AI in-house; we set it up and support your IT team with updates and model maintenance.
Before you buy, you test it free of charge for five days on our environment with your own documents and get a report of measured values.
Initial meeting: 30 minutes. Then we plan your five test days.
Shadow AI: your people already use AI, but not through your company
A ban rarely holds. Anyone facing a deadline pastes the paragraph into a browser, and your documents end up with a service your company has no contract with.
of the knowledge workers surveyed who use AI at work bring their own tools for it.
Microsoft Work Trend Index 2024, survey of knowledge workers
of the volume of corporate data entered into AI tools was detected as sensitive, up from 10.7% a year earlier.
Cyberhaven, March 2024, usage data from around three million employees
Sovereign AI gives your staff an assistant that works like the one in the browser, while the documents stay in-house. That lets your IT team close the route to outside services without slowing the work down.
What your staff do with it
The interface is a chat that most people know. Behind it sits what daily work needs.
Draft texts
Emails, proposals, statements and minutes as drafts, in your company's tone.
Translate
German and English in both directions, technical texts included.
Work through long documents
Several hundred pages in one pass: tenders, standards, manuals. Questions across several documents at once.
Read images and scans
Text recognition for scanned letters, delivery notes and photographed documents.
Output in your own layout
Results as PDF, Word or PowerPoint. Store your company template and the output carries your corporate design.
Assistants per team
Sales works with a different assistant than engineering, each with its own instructions and documents.
Web search with sources
If you want it, the AI searches the web and names its sources. Whether it may do so is your call.
Connection to your systems
Access to mail, document storage or ERP is possible, always with the rights of the person asking. We assess it with you once the core service is running.
What the sovereign environment gives you
Sovereign means you decide where the AI runs, who uses it and what of that use is documented.
Your requests stay in-house
- The model runs on hardware inside your network. What your staff ask the AI does not leave the company.
- Operation needs no connection to the model vendor.
- Web search is optional. Switched on, search queries go out to the web; switched off, everything stays in-house.
Predictable cost, replaceable models
- No billing per request. You plan hardware and operation, and the cost does not grow with every question.
- You are not tied to a single vendor: models can be replaced, and a second one can be added.
- We test new models on your test environment before they go into use.
Need-to-know access
- Sign-in through your directory, such as Active Directory, with two-factor authentication.
- Data rooms per team or department. Anyone outside a room does not see the documents in it.
- No admin view into your staff's conversation histories.
Traceable with OdyControl
- OdyControl documents usage: who used the AI, when and to what extent.
- It records operational data, not content. You set the scope according to your own rules.
- That lets you show auditors how the AI is used without recording conversations.
Four decision makers, four questions
An AI for every member of staff is rarely one person's decision, and each role brings its own question.
What does it cost, and what do we get for it?
Predictable cost for hardware and operation instead of fees per request. Before any investment comes the demo, with measured values from your own test.
Where does our data go?
Nowhere outside your network; only web search, if you switch it on, sends search queries out. The environment becomes part of your ISMS, with an operations manual and documented usage.
How much work does running it mean for us?
The hardware arrives preconfigured and tested, and sign-in runs through your directory. We take care of updates, model testing and monitoring. We only get access to the environment once you have agreed to it.
Is this used to monitor performance?
Usage is documented without content, to the extent you agree on together. Admin rights give no view into conversation histories either. We bring a template for the works agreement.
What the law, auditors and the works council expect
Running your own AI does not change your obligations. It makes them easier to meet.
- EU AI ActStaff need AI literacy (Article 4 of the AI Act). Certain AI systems also carry transparency obligations (Article 50 of the AI Act); we work out with you which of them apply to your use. We bring the documents your implementation builds on.
- Template for the works councilA draft works agreement: what is documented, who may see it and how long it is kept.
- Part of your ISMSThe environment is added to your ISMS under ISO/IEC 27001, with its risks and measures.
- Operations manualUpdates, backup, recovery and incident handling are written down before the pilot starts.
- Our accessWe only access the environment after agreeing it with your IT team. Where your company requires it, our staff can undergo security vetting under the German Security Clearance Act (SÜG).
From test to operation in five steps
Each step ends with your decision on whether to continue.
- 1
Initial meeting
We clarify what your staff should use the AI for, with which documents and under which requirements of your company.
You getA record of your requirements and a proposal for the demo.
- 2
Demo: five days of free testing
You work for five days on our sovereign environment, with your own documents that are public or cleared for the test. Every participant gets a personal login with two-factor authentication. We delete test documents and prompts every day.
You getA final report with measured values: response times, throughput, usage.
- 3
Pilot
The hardware sits in your network, preconfigured and tested by us. You start with key users and expand step by step.
You getA ready-to-run environment with sign-in through your directory, and the operations manual.
- 4
Production
Roll-out to all staff. The move happens without interruption, and accounts, histories and data rooms carry over.
You getThe AI for your whole company, with assistants and data rooms per team.
- 5
Operation
We install updates, test new models before they go live and take care of monitoring. The pilot hardware then becomes your test environment for new models.
You getOngoing operation by the manual, with a named contact on our side.
What the solution is, and what it is not
- The leading models of the large cloud providers are not available for self-hosting. We use open models at the current state of the art.
- In German these models come close to the large cloud models, and they are strong with long documents.
- We have shown that the models run in our operation without any loss of quality compared with the original.
- Figures on response times and throughput come from your own demo, not from a brochure.
How the offer relates to OdyAgent and OdyControl
Sovereign AI follows the same principles as our products: where it runs is a decision, and every use leaves a trace.
OdyAgent
The agentic framework: agents handle tasks from OdySecure in a capsule without internet access, on a language model we run ourselves or on yours.
OdyControl
The AI control layer: decides which model gets which task, and documents usage.
Frequently asked questions
What is sovereign AI?
Sovereign AI is an AI assistant that a company runs on its own hardware inside its own network. The company decides which model runs, who uses it and what is documented. Requests do not go to a model vendor, and there is no billing per request.
Which companies is sovereign AI suited for?
For companies whose documents must not leave the building: mid-sized firms with engineering and customer data, industry, defence suppliers and regulated sectors such as finance, healthcare and energy. Cybervize sets up the sovereign AI there and supports the IT team in operation.
Is this an on-premises AI?
Yes. The AI runs locally on hardware inside your company's network, not in someone else's cloud. That makes it a private AI for your company: you decide on the model, access and logging.
Does any request leave our network?
Not in operation on your hardware, as long as web search is switched off. The model runs inside your network, and operation needs no connection to the model vendor. The exception is the demo: it runs on our environment, so you test there only with documents that are public or cleared for the purpose, and we delete them every day.
Which models do you use?
Open models at the current state of the art, chosen for German and for long documents. We discuss the specific models in the initial meeting. New models are tested on the test environment before they go into use.
Is it as good as the well-known cloud assistants?
In German, current open models come close, and they are strong with long documents. The leading models of the large providers are not available for self-hosting. The demo with your documents shows whether it is enough for your tasks.
Can IT or management read along?
No. There is no admin view into conversation histories. OdyControl documents usage as operational data without content, to the extent you agree with the works council.
What do we need for the demo?
A few people who work with it for five days, and documents that are public or cleared for the test. The demo is free of charge. We set up the logins, each one personal and with two-factor authentication, and we delete test documents and prompts every day.
Who runs the environment?
Your company, on its own hardware inside your network. We support your IT team: we install updates, test new models and take care of monitoring. We only access the environment after agreeing it with you, and the operations manual says who does what.
Does this help against shadow AI?
Yes, if the assistant covers daily work. People with a convenient route in-house no longer need the browser; the demo with your tasks shows whether it is enough. Your IT team blocks direct access to outside services; OdyControl governs and documents the model calls that run through your own environment.
What does it cost?
That depends on the number of staff and the scale of the roll-out, and becomes specific after the initial meeting. There is no billing per request: you plan hardware and operation.
Five days with your own documents
In the initial meeting we clarify what you want to test, and then we set up the demo.
Your contact: Alexander Busse