A tenant reports a dripping tap. Before anyone can act, the team has to identify the flat, check the available records and ask for missing details. This preparation is a useful starting point for AI in property management: classify incoming requests, gather approved property information and prepare a case someone can work on. The responsible person decides on urgency, commitments and commissioning work.
The repair request below is an illustrative workflow. Its aim is a case a colleague can pick up without copying the same email into several systems again.
1. Match the message to the right flat
The message contains an address, a description and perhaps a photo. The system proposes a “repair request” category and compares the details with accessible property records. The draft includes the original message, contact, property and flat. An address matching several units is not enough: the flat stays unresolved and a question is prepared.
Before processing the request further, check whether a case already exists for the same issue. A follow-up message should add to that case instead of starting a second job. Your maintained property-to-team assignments determine who is responsible.
2. Add relevant records with their sources
An internal assistant can find approved instructions and relevant property information. It includes the source so the person handling the case can check what the proposal rests on. Not every document belongs in every case: access permissions determine which information is available for this task.
If a quotation later arrives as a PDF, document processing can extract the address, quote reference and description of the work, compare them with the case and suggest where it belongs. A conflicting amount or illegible flat number stays flagged for review. Extraction and matching prepare the work; the responsible person decides whether the quotation fits and should be approved.
3. Ask for the missing detail
If the flat is known but the report is too brief, the system prepares a specific question, such as which tap is affected. The tenant does not need to repeat information already available. A colleague sees the draft alongside the conversation and decides whether to send it.
This handoff matters for customer service. Whoever takes over needs the request, sources, open questions and previous contact together. Unclear or potentially critical reports go directly to a responsible person through the agreed process; an automatically assigned category does not replace an assessment of urgency.
4. Prepare an action someone can take over
| Case field | What the colleague receives |
|---|---|
| Request and match | Original message, contact, property and flat; uncertain matches stay flagged. |
| Sources and history | Relevant approved records and earlier messages; existing reports are linked to the same case. |
| Missing detail | Which tap is affected? The specific follow-up question is ready as a draft. |
| Next decision | The responsible person assesses urgency and ownership, then approves a question, routing or commissioning work. |
After review, the agreed workflow can carry out the approved step in a connected system. Whether your property software supports this depends on its actual interfaces and needs to be checked.
New colleagues can use the same knowledge base to ask which steps are required before commissioning work and open the current instructions. This makes onboarding part of the same workflow. Instructions, responsibilities and sources still need maintaining: AI does not make an outdated procedure correct.
Assistant, fixed automation or agent?
An internal AI assistant helps find information and answer questions: “What do we know about this property?” Fixed automation performs a predefined step, such as passing a reviewed case to the responsible team. When triggers and rules are clear, that connection may be enough.
An agent can connect several permitted steps based on intermediate results: retrieve information, identify a gap and prepare a draft, for example. It needs limited access, clear stop conditions and a person to handle exceptions. Our AI process automation service explains how we integrate and support these workflows. Start with the scope that reliably prepares your first case.
What our work with a property manager already shows
The published Kirsch & Drechsler case study covers workshops, internal knowledge management, a website chatbot, ongoing support and joint CherryGate development. It shows related experience building and operating AI applications. The repair workflow described here is an illustrative use case, not a claimed result from that customer project.
Make the first test easy to judge
Choose one common, manageable request type. Gather anonymised sample messages and related documents, including an ambiguous case and a duplicate report. List the systems involved, the responsible person and the desired result: for example, a correctly assigned draft with sources and open questions.
Have your team compare the drafts with their current work: Is the flat correct? Are the sources current? Is the next step permitted and understandable? Include corrections and rework in the comparison. This shows whether the workflow helps and which step is worth connecting next. We can discuss that first property workflow together.
