swiss digital services + AI
How we use AI in our services
Swisswise GmbH, the company behind MATE, carries the swiss digital services + AI label. This page discloses how AI is used in our services for business customers: what role it plays in each area, which models we use, where they run and what happens to your data.
Consulting and planning
- The role of AI
- In consulting we plan the use of AI: which processes it can support, with which models and in which operating mode. The analysis of your processes and the concept are done by people on our team.
- Hardware
- When we recommend hardware, it is for running open language models on your premises. The sizing depends on model size, context length and the number of concurrent users.
Building it in
- The role of AI
- Here AI is the key technology as the label defines it: the AI functions we build in do not work without it.
- Models
- We use state-of-the-art open-weight or open-source models, including models from the Qwen, Gemma, GLM and Apertus families. Which models run in a given case depends on the deployment. We curate, test and modify the models before use.
- Third-party services
- We do not use third-party services for model inference. External AI services are connected in a project only if you explicitly approve them.
- Your data
- Your documents and databases are indexed so the model can answer from them. The method is called retrieval-augmented generation. The model is not trained on your data in the process.
- Where it runs
- In our own data centres in Switzerland or on your equipment.
On your premises
- The role of AI
- The models run on your own hardware in your server room. Which ones depends on your use case and your equipment.
- Model maintenance
- We check new model versions before we offer them to you.
Confidential
- Training
- We do not use customer data to train models. That holds in the MATE Cloud and on your equipment alike.
The association's five dimensions
In integration and operation AI is the key technology. In consulting it is the subject of planning. The language models we use are open-weight models, and none of them is a third-party service.
- Hosting
- Self-hosted, in our own data centres in Switzerland or on the customer's equipment.
- Level
- Base models, run directly. No embedded service that model inference would depend on.
- AI technologies
- Large language models, generative AI and retrieval-augmented generation.
- Origin and type
- Open-weight or open-source models, including models from the Qwen, Gemma, GLM and Apertus families. No closed-source models from third parties.
- Manipulation and fine-tuning
- We curate, test and modify the models before use.
These services are provided by Swisswise GmbH. The company discloses its use of AI on its own site as well.
The services disclosure at swisswise.comQuestions about this disclosure
If anything is unclear, write to us.
Contact