AI in property management: three workflows that ease the shared inbox

Starting point: more correspondence than people to write it

Digitalisation in the Swiss construction and real estate sector is moving in small steps. For the Digital Real Estate Index 2026 by pom+ Consulting, around 200 specialists and executives were surveyed. Property management reaches 4.4 out of 10 points, the highest value since the survey began. The overall index stands at 4.3. In the hype cycle, AI and machine learning are the only one of the twelve technologies surveyed that show clear progress. Around two thirds of respondents see high value in them, down from 75% in the previous survey. What slows the development down are questions of investment and cost.¹

On top of that comes the need for staff in day-to-day operations. SVIT notes that a lack of experienced personnel in facility management gets in the way of maintenance and of the smooth operation of properties.² Whoever cannot fill an open position spreads the work across the existing team. The first thing to suffer is the response time towards tenants, and an unanswered repair report turns into a complaint to the owner two weeks later.

A large part of that work is correspondence that follows a fixed pattern: the service charge statement and the questions that follow it, the rent adjustment tied to the reference interest rate, the formal requirements for terminations, the handover report with the orders to the contractors. Then there is the language mix. The same matter goes out in German, French, Italian or English depending on the tenant. So the question is no longer whether a management office uses AI. The question is whether the AI works with the documents of its own properties, and where the tenant data sits while it does.

Use case 1: sorting and answering tenant enquiries

On Monday morning the shared inbox holds the messages from the weekend: a dripping boiler, questions about the service charge statement, a noise complaint from the third floor, a termination for the end of the quarter, several requests for a spare key. The AI colleague reads the inbox through the connected mailbox. It assigns each message to a type of request and to a property, and places a draft reply next to it.

The draft comes from the documents of the property concerned: the tenancy agreement, the house rules, the regulations and the previous correspondence with this tenant. For the repair report, the draft states who fixes the fault and what information is still missing. For the spare key request, it states the rule from the regulations, with the source reference, so the statement can be checked in seconds. The message goes to the responsible property manager, who reads it, corrects it and approves it.

What matters in practice: no new tenant portal and no second system appear. The AI colleague works in the mailbox you already have, and tenants keep writing to the same address. The red lines stay in place too: terminations, rent adjustments and formal warnings are legally binding notices and never go out automatically. And a repair report stays a report, whatever the draft makes of it: under Article 257g of the Code of Obligations, tenants must report defects they do not have to remedy themselves, and once it arrives the landlord is on notice.⁶ The formal requirements and deadlines of Swiss tenancy law remain the responsibility of the management office.

Use case 2: explaining the service charge statement in plain words

After the statements go out, the same wave arrives every year: why is my statement higher than last year. The questions all land within the same two weeks, and they land on a team that is processing the additional payments and the refunds at the same time.

The AI colleague pulls up the statement for that apartment, the allocation key and the figures from the previous year, and drafts an explanation in plain language: which item has gone up, how the share is calculated, what the comparison with last year shows. The draft is written in the language the enquiry came in, in German, French, Italian or English. The property manager checks the figures and approves. That makes the seasonal peak manageable. It no longer has to be absorbed through temporary staff alone.

What the workflow does not do: it does not correct a statement and it does not charge anything. It explains what the statement says. If an item really is wrong, the case goes to accounting, as before. If a tenant asks to inspect the receipts, that is not a special favour but a right under Article 257b paragraph 2 of the Code of Obligations, and the office grants it.⁶ Because every explanation rests on the stored statement, the property manager sees immediately what the draft is based on.

Use case 3: from the handover to the contractor orders

The handover takes an hour, the paperwork afterwards often takes longer. On the way back the property manager dictates the notes into the phone, the device’s dictation function is enough, or types them in: kitchen front scratched, silicone in the bathroom to be renewed, living room walls to be painted, photo list attached.

Out of that come the draft handover report in the company format and the worded orders to the painter, the plumber and the cleaning company, with address, apartment number, access arrangements and deadline. On the same day, instead of at the weekend. Who gets the job, who bears the cost and where normal wear ends and damage begins is decided by the management office. The AI colleague supplies the text.

Here too: no order goes out without approval, and nothing is charged automatically. What is billed to the tenant is decided by the office, and every change to a draft stays traceable.

Connectors: what can be linked to GARAIO REM, Abacus and the rest of your software

An AI colleague is only as useful as what it is allowed to reach. MCP is an open standard for connecting external systems. Through it, programs with a documented interface can be addressed. An interface, often called an API, is a described route along which two programs exchange data. Permissions apply per individual tool, and every call is written to the log. This is the state of play for common property management software, checked in September 2026:

Is your software missing from the list? Much of the daily work runs on the mailbox and on documents and needs no system connection at all: sorting the enquiries, the service charge explanation, the handover report. How a connection via MCP works in principle, and which systems are prepared, is described on the integrations page.

Data protection: tenant files are sensitive

A tenant file holds more than a name and an address: debt collection register extracts, credit reports, proof of income, sometimes a social assistance confirmation or a medical certificate. Most of that is ordinary personal data. Information about social assistance measures and about health, by contrast, counts as sensitive personal data under Article 5 of the Data Protection Act.⁷ Responsibility stays with the management office in any case, even when an AI tool processes the data. If a provider processes the data on its behalf, Article 9 of the Data Protection Act applies: the provider may only do what the office itself would be allowed to do, it has to ensure data security, and it may pass work on to a subcontractor only with prior approval.⁷ This article is not legal advice.

On top of that come the expectations of the owners. Institutional owners such as pension funds and insurers regularly require in their management contracts that the data of their properties is processed in Switzerland. Kisuno runs in data centres of Green in Lupfig that are certified to ISO 27001, or on your own server.

Customer data does not train any models, and permissions apply down to individual tools. Every access is recorded in the audit log, a record you can export at any time. The data processing agreement and the subprocessor list are published openly on the website. One thing belongs in the same answer: on offer are the current models from OpenAI (ChatGPT) and Anthropic (Claude); they run in the AWS data centre in Frankfurt. AWS is a US company, so the US CLOUD Act applies even with servers in Frankfurt, and we say so openly. For customers with higher data-protection requirements, open frontier models such as GLM, Kimi and DeepSeek are available, operated by European companies on servers in Europe; the CLOUD Act does not apply there. Anyone who has to rule that out runs the model and the instance themselves. When an owner asks where the data of their properties is held, you can answer both points in one sentence.

Frequently asked questions from property management

Sources

All figures and interface details in this article come from the following sources, last checked in September 2026.

1. pom+ Consulting, Digital Real Estate Index 2026, media release of 9 March 2026, around 200 specialists and executives surveyed: pom.ch
2. SVIT, Real Estate Symposium: shortage of qualified staff in the Swiss real estate sector, 12 January 2024: realestatesymposium.ch
3. GARAIO REM, developer guide with the GraphQL interface and the message bus: github.com
4. W&W Immo Informatik, documentation of the ImmoTop2 DMS interface: github.com
5. SwissRETS, open standard for property data: github.com
6. Swiss Code of Obligations, Articles 257b, 257g, 266c, 266l, 269d and 273, wording in the Classified Compilation of Federal Legislation: fedlex.admin.ch
7. Swiss Data Protection Act, Articles 5 and 9, wording in the Classified Compilation of Federal Legislation: fedlex.admin.ch