← HomeApproach and ways of working
Approach and ways of working
Approach
No ready-made recipe
AI moves faster than any method tied to one state of the art. So we work iteratively: build on what exists, remove what is no longer needed, merge what overlaps, and add new capabilities as an extra layer.
Our own practice
Our working environment has been in daily use since spring 2025 and has been fundamentally reworked several times since. It has three areas kept apart: daily operations, client work and a lab. New methods are tried in the lab and adopted into operations only once they have proven themselves.
Minimally invasive, not a big project
Your systems and processes are the starting point. We first get a precise overview, then work on a few deliberately chosen points and change things only where it demonstrably pays off. No change of existing systems and no large migrations are needed for this.
Technical and organisational principles
| Principle | Implementation |
|---|---|
| Data | The trial starts with freely licensed or synthetic material; your own data only after clarification with your data protection officer, then processed locally |
| Processing | EU data centres with a data processing agreement, or fully local |
| Approvals | Results are checked and approved by people before further use |
| Separation | The trial environment is separate from your production systems |
| Traceability | We document what ran with which data and who approved it |
| Evaluation | Methods are measured on real tasks before they are introduced |
| Openness | Open standards and open-weight models, no tie to one vendor |
In your environment
For the step after the trial we specify a lab environment in your infrastructure or with a provider of your choice, with configuration, evaluation method and documentation. Operations and the decision stay with your IT.