Hüni — KI-Gemeindeassistent.

Residents often spend time searching for the right information or the appropriate online service on municipal websites. Although Hünenberg already provides a modern online service portal and digital offerings such as the eZug app, information retrieval and navigation remain a key part of many citizen interactions.

Conventional AI chatbots are only partially suitable for use in the public sector. Responses may be inaccurate when the underlying data is incomplete, data processing often takes place outside Switzerland, and sources are either missing or difficult to verify. Traditional FAQ systems, on the other hand, quickly reach their limits when faced with the diversity of individual citizen inquiries.

Hüni is an AI-powered municipal assistant built on Retrieval-Augmented Generation (RAG). Its knowledge base consists exclusively of verified and publicly available municipal information sources, including regulations, official publications, online services, forms, and other relevant content.

The system is based on three core principles: traceable source references for informational responses, no handling of personal administrative cases, and consistent refusal to answer when sufficient information is unavailable. If a question cannot be answered reliably, the user is referred to the appropriate municipal office.

The architecture is designed entirely around Swiss data sovereignty. Hosting and data processing take place in Switzerland and can be operated in compliance with Swiss data protection requirements. Estimated operating costs for a municipality with approximately 9,000 residents are between CHF 200 and CHF 400 per month.

Primary Target Groups

• Small and medium-sized Swiss municipalities
• Cantonal administrations focused on digital citizen services
• Municipal offices with a high volume of recurring information requests

Secondary Target Groups

• Municipalities with multilingual populations
• Municipal associations and shared-service organizations
• Providers of digital government solutions as an integrable white-label component

Q2 2026 — Pilot

Functional prototype with a curated knowledge base (≈30 entries), browser-based widget, and offline mirror. Execution of response quality and robustness testing using real-world use cases. Evaluation by interested municipal stakeholders.

Q3 2026 — Backend

Cloudflare Worker as an API proxy. Vector database (Qdrant, Swiss hosting). Sitemap crawler for automated content ingestion. Planned user and field testing with residents.

Q4 2026 — Pilot Operation

Preparation of a production-ready pilot operation in collaboration with an interested municipality. Analytics dashboard for administrative staff (frequently asked questions, escalations, usage patterns). Continuous refinement of the knowledge base.

2027 — Scaling

Assessment of transferability to additional Swiss municipalities. Full multilingual support (DE / FR / IT / EN). Evaluation of integration with existing digital identity solutions such as eZug.

Scientific Foundation

• Lewis, P. et al. (2020): Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv:2005.11401.

Data Protection & Governance

• Swiss Federal Data Protection and Information Commissioner (FDPIC): Guidelines for the Use of Artificial Intelligence in Public Administration, 2024.
• OECD (2023): Recommendation on Artificial Intelligence – Public Sector Application.

Swiss Digitalization Strategies

• Swiss Federal Chancellery: Digital Administration Switzerland Strategy 2024–2027.
• Canton of Zug: eZug – Digital Administration, strategy documents and public information.