Residential Property Management and Responsive Repairs: Can AI Replace Catalogue-Based Maintenance Troubleshooting Tools at Large German Housing Companies?

Abstract

In repair management at large German housing companies, defect catalogues help structure tenant reports, standardize dispatch decisions, and support compliance with liability and operatorresponsibility requirements. This paper examines whether large language model (LLM) systems can replace such catalogue-based tools in responsive repairs. The study combines qualitative analysis of repair tickets, catalogue structures, and process flows with exploratory LLM prototypes and expert interviews across large housing companies, technical service providers, and smaller landlords. The prototypes were tested on repair descriptions and compared with existing catalogue-guided workflows. Without catalogue access, LLMs often produced plausible outputs for frequent, low-risk cases but showed uneven classifications and emergency priorities in rare, safety-critical situations such as suspected gas leaks. Catalogue-guided flows were more stable, auditable, and better aligned with legal and operational requirements. At the same time, LLMs proved useful as assistive tools for summarizing tenant messages, generating follow-up questions, detecting duplicates, supporting rule checks, and improving documentation. The paper concludes that full replacement of catalogue-based troubleshooting is not yet realistic; near-term potential lies in hybrid systems that combine defect catalogues with carefully governed AI support.