AI Cube. Local AI for your company.
The AI Cube is a preconfigured AI server that sits in your network. Employees query your own company knowledge through a chat interface in the browser. Inputs, documents and answers stay in-house.
Local, hybrid or dedicated in the EU · Set up and supported from a single source
Who the AI Cube is for
For companies that want to use AI with confidential data: engineering documents, contracts, personnel data, customer projects. With a cloud service, this data would have to go to a provider. With the AI Cube, model calls and approved knowledge sources run in your own environment.
Local operation does not automatically make a process GDPR-compliant. Purpose, legal basis and permissions still have to be right. But there are fewer places data flows to, which makes the assessment easier.
What is inside the AI Cube
Language models
Suitable open or licensed models, selected by language, quality, speed, licence and hardware requirements. We test any model change in advance with reference questions.
Search in company knowledge (RAG)
Documents from document management, file servers, Microsoft 365 or specialist systems are made searchable. Answers cite the source and respect existing access rights.
Chat interface
The browser interface works like a familiar AI chat and is connected to your user accounts and groups.
Controlled agents
Defined tools with which the AI combines information or prepares drafts. Permissions, approvals and logging limit what it is allowed to execute.
Hardware
Computing power matched to the number of users and the models, delivered preconfigured and integrated into your network, identities and data sources.
What companies use the AI Cube for
Technical knowledge
Search manuals, service reports and specifications, with answers that cite the source.
Internal rules and processes
Make work instructions, policies and process knowledge accessible for each role.
Quotes and service
Combine information from the knowledge base and CRM, prepare draft replies and quotes.
Document review
Classify documents, extract content and flag deviations for expert review.
Project knowledge
Make project documents, decisions and experience findable across departments.
Security and operation
Local operation does not replace a security concept. The AI Cube therefore includes role-based access, separate network segments and encrypted connections. Usage and agent actions are logged as far as data protection allows. We regularly test answer quality with reference questions. Updates, backups and recovery after an outage are planned.
We follow the EU AI Act, the BSI recommendations on generative AI and the NIST AI Risk Management Framework. In addition, ISO/IEC 42001 and the security guidance of the OWASP GenAI Security Project.
When local operation is not worth it
For tasks without confidential data, a language model from a provider with a data centre in the EU is usually cheaper. Many solutions therefore combine both: sensitive steps run on the AI Cube, everything else in the EU environment.
Whether your own hardware pays off depends on the number of users and the volume of data. We clarify this before any purchase.
Rollout in three steps
For the AI Cube to be used, rolling it out among employees is part of the job.
Analysis
Clarify the goal, user groups, data sources, protection requirements and measurable success criteria.
Pilot
Connect a limited knowledge base, test the quality of answers and review them with real users.
Rollout and operation
Connect further sources and groups, train teams, appoint key users and monitor usage.
Frequently asked questions about the AI Cube
Does data leave the company with the AI Cube?
In fully local operation, model calls and approved knowledge sources stay in your environment. Whether data goes to external services depends on connected interfaces, telemetry and configuration. We define and document these data flows in the project.
Which language models run on the AI Cube?
Suitable open or licensed models, selected by use case, language, quality, speed, hardware requirements and licence. Before every change, we test with reference questions.
Is local AI automatically GDPR-compliant?
No. Where it runs is an important building block, but it does not replace reviewing the purpose, legal basis, data minimisation, transparency and permissions for the specific use case.
How reliable are the answers from company knowledge?
The AI shows the sources, which makes answers verifiable. This does not prevent errors completely. Quality depends on the documents, the search and the model. Critical results are checked by a qualified person.
Does the AI Cube adopt our existing permissions?
Yes, provided the connected sources support it. User accounts, groups and access rules feed into the permission concept, so that everyone only sees what they are allowed to open today.
What is the difference between the AI Cube and Microsoft 365 Copilot?
Copilot runs in the Microsoft cloud and works with your Microsoft 365 data. The AI Cube runs in your own network and can also include sources outside Microsoft 365. Which solution fits depends on your protection requirements and your existing software.
Assess local AI for your company
In the free initial consultation, we clarify which data the AI Cube should make accessible and whether local operation is the right approach for you.
