Zurich canton says its second round of AI sandbox trials shows that artificial intelligence can deliver real value when used for clearly defined tasks with clean data, human oversight, and a clear legal framework. The projects ranged from building permits and medical documentation to a digital eye clinic and bridge-monitoring and inspection robots.
The second phase of the AI sandbox shows that artificial intelligence creates added value primarily where it supports concrete, practical tasks, said Raphael von Thiessen, Head of the AI Hub Program at the Office for Economic Affairs.
“Therefore, widespread use requires not only high-performance models, but also test environments that combine the technology with clean data, clear processes, human oversight, and reliable legal frameworks,” said von Thiessen.
Carmen Walker Späh, the cantonal economy director and president of the government council, said Zurich wants to use AI where it can “provide tangible relief for people.” She said the building permit project shows how responsibly deployed AI can improve data quality and efficiency without weakening the authorities’ decision-making power.
Read More: AI v/s Defense: Is Switzerland’s New Cyber Strategy Actually Working? – Helvetica Times
Zurich Tests AI Deployment In sectors where routine tasks are costly and repetitive
The most practical example came from the building permit process, one of the most paperwork-heavy and time-consuming areas of cantonal administration. Working with Gossweiler Ingenieure, Byte Studio and Nokema, the canton developed a prototype for an AI-supported preliminary check in the notification procedure, such as for solar panels, heat pumps and charging stations. The system can flag missing or incomplete documents early, but it does not assess the substance of an application, that remains the job of the authorities.
The project was based on real permit data from the city of Kloten and carried out in close coordination with the cantonal building department. Zurich said the results should help it respond to a parliamentary motion on how to ease pressure on building authorities and speed up approvals.
The canton’s Office for Economic Affairs has been running the AI Innovation Sandbox since 2022 with partners from administration, business and research. In the second phase, 24 organisations submitted proposals and five were selected for testing under real-world conditions.
Beyond building permits, the other projects pointed to a wider pattern of AI deployment in sectors where routine tasks are costly and repetitive.
In medical documentation, the canton and MPAssist tested whether AI could help doctors transcribe and structure reports, reducing administrative burden while still meeting strict data-protection requirements. In the digital eye clinic project, Zurich Triemli Hospital and the Spross Foundation examined whether locally run AI models could reliably analyse retinal images in practice.
Another project looked at sensor-based bridge monitoring for a railway bridge in Wädenswil, in cooperation with Irmos technologies and the Südostbahn. Zurich said the findings showed sensor data can complement traditional inspections and support safer, more sustainable maintenance decisions for aging infrastructure.
A fifth project, with Zurich-based robotics company ANYbotics, explored the regulatory conditions for autonomous inspection robots in critical infrastructure such as energy facilities.
The canton also said the sandbox is generating findings that matter beyond individual pilot projects, including for companies, public bodies and researchers working on responsible AI use. It added that the work feeds into national discussions on AI regulation and related measures.
Euro Area Unemployment Remains Unchanged At 6.3%
Third Phase Of AI Sandbox Has Already Begun
Zurich has already begun evaluating applications for the third phase of the sandbox. With 57 submissions, interest has risen sharply from the previous round, and most came from health, core administration, and construction and spatial planning. The canton said it also received several proposals focused on AI governance, suggesting strong demand for clearer rules on responsible use.
