AI for everyone. On the company's terms.

AI tools had arrived in the company, but everyone used them differently: without access to the company's knowledge, without shared rules and without any overview of which data was going where.

At the same time, a great deal of knowledge was stuck in people's heads, folders and emails. Anyone who needed an answer had to know whom to ask.

AI for all employees, structured and suited to the way the company works: an assistant that knows what the company knows, works with its own systems and only sees what each user's role permits.

It began with a target vision: what are the use cases, which systems are involved, which rules apply? A fundamental decision was not to tie the company to a single AI provider. That is why an open standard for integration, the Model Context Protocol, and clearly described processes form the foundation.

On this basis we built the knowledge base and the server, with the first knowledge areas and processes. Since then the platform has grown step by step, one use case at a time.

• One point of access for everyone. Employees work with Claude as their assistant. A dedicated MCP server connects it to the company.
• Sign-in with Microsoft 365. Access rights follow the company's Microsoft 365 groups. Everyone sees only the knowledge and functions their role permits.
• Structured knowledge base. Company knowledge and processes sit in a clearly organised topic tree, with fixed rules for structure and expansion. A hybrid search finds entries by keyword and by meaning.
• Maintaining knowledge in the chat. Employees suggest additions directly in the conversation. Those responsible approve them, after which the new version is available to everyone. Nobody needs to know any technical tools to do this.
• Integration of in-house software. The assistant works with the company's own project management software, for example when creating master data with a duplicate check.
• Filing email attachments. The assistant finds the email, uses the folder structure from the knowledge base to determine the right storage location in SharePoint, suggests it and files the attachments once confirmed.
• Commercial reports on demand. The assistant knows how the project data is structured and produces reports on request, without the need to build dedicated dashboards.
• Control built in. The assistant asks before every change. Every access is logged, and the platform runs on AWS in Frankfurt.

• Knowledge on demand. Employees ask in their own words and get answers drawn from verified company knowledge.
• Everyone sees only what they are allowed to see. The existing roles from Microsoft 365 also apply to the AI.
• Knowledge grows in everyday work. Additions are made where gaps are noticed and go live immediately once approved.
• Open to more. New systems are connected as additional tools, without rebuilding the assistant.