AI for Property Management: PROPbase Embraces Intelligent Process Automation

Key Takeaways

• Tenant requests and damage reports often cause a great deal of manual effort for property management companies.
• PROPbase wanted to make this process in digital property management more efficient.
• To achieve this, PROPbase and Evolit jointly developed the AI-powered claim agent as a new feature of the PROPbase platform.
• A multi-stage AI pipeline structures tenant requests and prepares them for further processing in the existing ticketing system.

Starting Point & Challenge: When Unstructured Tenant Requests Create Manual Work

Property management companies handle numerous tenant requests every day: water damage, broken appliances, heating problems, suspected mould or general complaints about a property. The problem? These reports usually arrive as informal emails and are rarely standardised. They contain typos, dialect, emotional wording or incomplete information.

To process them, however, it is not enough to understand the content only roughly. The message must reveal which property is affected, whether a contractor needs to be commissioned and which trade would be responsible.

For property managers, this process is very time-consuming and error-prone. Manually identifying the right trade and a suitable service provider creates effort, slows down communication and can lead to inconsistent processing quality.

As a software provider, PROPbase continuously develops its property management platform along the needs of its customers. For this use case, the existing platform was to be extended with an AI-powered claim agent that automatically analyses incoming emails, structures the relevant information and generates a proposal for further processing. For the design and development of this AI component, PROPbase partnered with Evolit as a specialised technology partner.

A purely rule-based automation would have been too rigid for this. The use case requires semantic text understanding: the solution has to interpret natural language, recognise ambiguities and combine business rules with property, tenant and contract partner data.

Analysis & Approach: From Use Case to Integrated AI Solution — Together

The project began with a thorough joint analysis. In several workshops, PROPbase and Evolit fleshed out the use case together and clarified how tenant requests are handled in practice, which information is relevant for making the right decision and where automation can create the greatest value.

This early clarification of the business domain was one of the success factors. Particularly in AI projects, it is not enough to simply “apply” a model to a process. It is essential to understand the process itself, its exceptions and its business dependencies in detail. Only then could a solution emerge that does not work generically but is tailored precisely to PROPbase’s requirements and the workflows of property management companies.

PROPbase contributed its industry, product and platform expertise as well as the requirements of property managers. Evolit complemented this with specialised AI and architecture expertise. Design, development and integration were closely coordinated between both teams and jointly refined over several iterations.

“For us, it was particularly valuable that Evolit always evaluated the technical possibilities from the perspective of our specific business process. The result was not a generic AI solution, but an individual solution that precisely fits our platform and the requirements of property management companies,” says Matteo Carosella, member of the executive board.

The jointly developed claim agent extends the existing PROPbase platform with multi-stage AI processing. The Python pipeline built for this purpose is integrated into the platform via a standardised interface and can connect different AI models through configuration.

Incoming tenant requests are first pre-processed, and sensitive information such as email addresses or phone numbers is anonymised. An LLM-based processing step then handles the domain classification: the solution assesses whether the message should result in a ticket or a contractor order, assigns the request to the matching property and derives a concrete processing proposal from it.

This goes far beyond simple text recognition. The pipeline combines semantic language understanding with existing PROPbase data, such as information on properties and tenants. In addition, customer-specific service instructions and historically qualified cases are used so that the AI can classify requests in the correct business context. In this way, an informal email becomes a structured proposal for further processing in the ticketing system.

The technical architecture was also deliberately designed for flexibility, data protection requirements and cost control. The solution was conceived to fit into the existing PROPbase environment, so no additional external solution had to be integrated into the sensitive core process of tenant requests. The LLM connection is implemented provider-independently, allowing PROPbase to use, compare and, if necessary, switch between different models. This makes it possible to balance quality, speed and cost as required, without having to fundamentally redevelop the solution.

Transparency about AI usage was also considered for later operations. The solution makes it traceable how many AI resources are used per request, creating a basis for cost control at high transaction volumes. At the same time, the claim agent is designed as a modular service within the PROPbase architecture and can therefore be flexibly integrated into existing platform processes.

Results & Benefits: Faster Tenant Communication and Better Cost Control

The PROPbase claim agent lays the foundation for classifying tenant requests with consistent quality and keeping running costs under control even at high transaction volumes.

A key advantage of the custom development lies in its integration into the existing PROPbase infrastructure. The solution was designed to run within the existing platform and data processing workflows. As a result, PROPbase did not have to integrate an additional external SaaS product into the sensitive core process of tenant requests. This reduced the coordination effort around data protection, approvals and existing customer agreements, making the solution considerably easier to implement organisationally as well.

PROPbase thus gains a differentiating feature for a central process of its platform: tenant requests and damage reports can be automatically pre-qualified, classified by domain and processed further as structured tickets. Property managers are relieved of manual pre-screening, the selection of the right trade becomes more consistent, and communication with tenants and technicians can happen faster.

“Evolit understood not only the technical side of our use case, but above all the business process behind it. Through the joint analysis, we were able to develop an AI solution that precisely fits our platform, our data and the requirements of property management companies,” says Matteo Carosella, member of the executive board.

Economically, the high transaction volume is particularly relevant. Instead of ongoing per-request fees for external SaaS providers, PROPbase relies on an individual solution with controllable operating and usage costs. The key metrics in focus are classification quality and the average cost per processed email.

The project shows how PROPbase, together with Evolit, is deliberately extending its platform with AI capabilities: not through an isolated tool, but through an integrated and maintainable solution that brings together product knowledge, domain logic and AI expertise.

About PROPbase

PROPbase is a Swiss all-in-one solution for digital property management that combines ERP, CRM, document management, ticketing and accounting in a single web-based platform. From small teams to large property management firms, PROPbase supports managers, owners and trustees in making property management smarter, faster and more transparent. Learn more at propbase.ch.