Future Fire Protection with Google Cloud AI Agents.
Summary
- Innovation through Collaboration: The Aargauische Gebäudeversicherung (AGV) partnered with building insurance companies from Vaud, Zurich, and Bern to develop a cutting-edge, AI-supported system to address the complexity of 2026 Swiss fire safety regulations.
- Technological Hybrid Approach: The project with PCG successfully implemented a Proof of Concept using a hybrid architecture that combines Google’s Gemini Enterprise Agent Platform with deterministic rule evaluation. This ensures both natural language understanding and precise compliance checks.
- Proven Efficiency Gains: The pilot demonstrated significant improvements in professional workflows, including drastically reduced research time for complex regulations and a guided multi-agent process that simplifies initial assessments for building applications.
About the Company
The Aargauische Gebäudeversicherung (AGV) is has been a reliable partner for the protection of buildings in the Canton of Aargau since 1805. As a public-law company, it does not pursue profit but is committed to the welfare of its insured parties. With its core areas of insurance, prevention, and intervention, AGV works to prevent damage and minimize financial consequences for building owners.
For this innovative AI project, AGV joined forces with the building insurance companies of the cantons of Vaud (ECA), Zurich (GVZ), and Bern (GVB) to jointly develop future-oriented solutions for fire protection in Switzerland.
Challenge and Goal
The complexity of the 2026 Swiss fire safety regulations places high demands on the precise evaluation and application of regulatory frameworks. The goal of AGV and its partners (VD, ZH, BE) was to validate the technical feasibility of an AI-supported system that can precisely process complex regulations. The aim was to radically accelerate access to structured fire safety information (BSV) while ensuring documentation creation results are as error-free as possible.
A particular focus of the project was to develop a guided dialogue through the new PoC application, enabling fire protection experts to gather all necessary information required for correct fire safety requirements. The result is then a systematized report that serves as the basis for fire protection planning.
Solution
In collaboration with PCG, a Proof of Concept (PoC) for an AI-supported fire protection agent was implemented. The system is based on a hybrid approach that combines state-of-the-art Generative AI with deterministic rule evaluation.
- Gemini Enterprise Agent Platform (formerly Google Vertex AI): Gemini-2.5-Flash and Gemini-2.5-Pro are used as inference engines to understand natural language and generate precise answers.
- Agent Search: Enables high-performance search within a structured knowledge database for fire safety regulations.
- Multi-Agent System: Specialized expert agents are orchestrated via the Agent Development Kit (ADK) to guide users through complex review processes.
- Cloud Run & Firestore: These serverless services ensure the secure deployment of frontends and the encrypted storage of chat histories.
Collaboration with PCG
PCG supported AGV and its partners as a strategic technology partner in the technical validation of the hybrid system. Thanks to their extensive expertise in the Google Cloud Tech Stack, a functional prototype was realized in the shortest possible time, translating the theoretical requirements of the draft future fire safety regulations into a practical, digital application. PCG’s professional orchestration of the specialized agents enabled AGV and its partners to successfully implement their vision of modern, networked prevention work.
Result
The pilot phase with experts from four cantons provided impressive evidence of the system’s performance. By analyzing the sessions conducted, the following key advantages were identified:
- Accelerated Research: The AI agent reduces the search time in the complex fire safety regulations. Information that previously had to be manually reconciled with difficulty is now quickly available.
- Guided Expert Dialogues: Thanks to the multi-agent architecture, users are guided through complex review processes, and the initial assessment for building applications is simplified.
- Validated Technical Feasibility: The PoC has proven that the combination of Google Gemini Enterprise Agent Platform and specialized agents functions well enough for a tool for productive nationwide use to be pursued.
The implementation of AI agents has proven highly effective in streamlining fire protection regulatory compliance and research, validating the potential for productive, nationwide deployment.