AI use cases implemented in Switzerland: In the interview, Syrian Hadad and Lukas Hebeisen talk about data-based management, digital independence and the responsible use of AI.
How can a cantonal administration use machine learning and generative AI sensibly while complying with data protection regulations? In this interview, Syrian Hadad, Head of Technology and Solution Development at the Canton of Aargau, and Lukas Hebeisen, Head of Product Line Cloud at Swisscom, talk about the implementation of data-based administration and digital independence in the Canton of Aargau.
Mr Hadad, the canton of Aargau is considered a pioneer in the use of the Swiss AI Platform. What prompted you to take this step?
Syrian Hadad: A few years ago, we gained valuable experience in the field of machine learning. With the rise of large language models, we were faced with the challenge of enabling our employees to use generative AI without exposing sensitive administrative data. We needed a solution that would fulfil our data protection guidelines, integrate into our infrastructure and be flexible enough for different models.
The Swiss AI Platform offers exactly that. It provides standardised interfaces for documents that we translate into encrypted fingerprints (embeddings). We orchestrate these building blocks via Open WebUI and operate everything in the cantonal data centre. This means we remain data sovereign, save on licence costs and can use the right model for each department. This modular approach makes us fast, flexible and independent.
What role does data storage in Switzerland play in this project?
Syrian Hadad: Data storage in our own data centre is central. The cantonal data protection law requires that personal data is only processed under the control of public bodies. When our AI platform summarises texts or transcribes Swiss German, all documents remain within the Swiss legal system.
This safeguards our digital sovereignty: we retain control over infrastructure, access protocols and can change models independently. We can use pre-trained models from abroad – the decisive factor is that we orchestrate them independently. In this way, we use modern technology without having to train expensive models ourselves and remain independent of individual cloud providers.

“The real challenge lies not in the technology itself, but in its intelligent embedding in social processes.”
Syrian Hadad, Head of Technology and Solution Development at the Canton of Aargau
You show that it is possible to use state-of-the-art AI technology and fulfil the highest security and sovereignty requirements at the same time. The development of the Swiss AI Platform seems to be based on a similar understanding. Mr Hebeisen, what vision is Swisscom pursuing with the development of the Swiss AI Platform?
Lukas Hebeisen: We have built an AI infrastructure that is tailored to Swiss needs. With the Swiss AI Platform, we offer a modular system for secure AI applications with data storage in Switzerland.
It was important to us that the platform does not create any new dependencies, but rather strengthens our ability to act. Control over infrastructure, data and processes remains entirely with the users. From the outset, we have pursued the goal of maximum data sovereignty, with processing exclusively in Swiss data centres – a clear commitment to digital sovereignty.
What services does Swisscom offer on the Swiss AI Platform?
Lukas Hebeisen: The Swiss AI Platform provides a modular infrastructure with which companies and public institutions can securely develop, operate and integrate AI applications.
Specifically, we offer services in three areas:
First, we enable the secure operation of foundation models, including language models for text analysis, classification or summarisation – hosted entirely in Switzerland.
Second, we provide tools for data processing, for example for transcribing dialects, analysing documents or semantic searches.
And third, we support our customers in orchestrating and monitoring these services, focussing on transparency, logging and controllability – all in compliance with Swiss standards and data protection requirements.
Our aim is not to build everything ourselves, but to make the best available models accessible in a controlled manner – without relinquishing control of the data.
“Our aim is not to build everything ourselves, but to make the best available models accessible in a controlled manner – without relinquishing control of the data.”
Lukas Hebeisen, Head of Product Line Cloud, Swisscom

You show how an AI infrastructure can be set up strategically. What does this look like in practice? Mr Hadad, where do you actually use AI in the canton of Aargau?
Syrian Hadad: We use AI wherever it brings immediate benefits: In the forest, we use lidar images to create detailed maps for forestry vehicles. A voicebot at the road traffic office handles standard enquiries. A system automatically reads handwritten forms. At the cantonal police, Swiss German interviews are transcribed into written language. Soon AargauGPT will also be further optimising our processes.
How are your employees reacting to these changes?
Syrian Hadad: From an IT perspective, it’s tricky to speak for everyone. I mainly get feedback from the early adopters, and it’s consistently positive. AargauGPT has only been running in IT and a few pilot areas so far. The canton-wide deployment is accompanied by training courses and practical guidelines so that everyone experiences the same benefits.
How do you see the future of AI in Switzerland?
Lukas Hebeisen: Switzerland is ideally placed to play a pioneering role in the development of trustworthy AI: first-class education, political stability and a strong culture of innovation. Despite all the technological enthusiasm, it will be crucial that we do not lose sight of people and ethical principles.
Syrian Hadad: Exactly. The real challenge lies not in the technology itself, but in its intelligent embedding in social processes.
Swiss AI Platform
With the Swiss AI Platform, Swisscom offers a secure Swiss AI platform with comprehensive data protection. The customised solution for GenAI applications protects your data and gives you access to high-performance processing power for your AI calculations.