AI for property management: five processes to automate first

Part 01

Why AI for property management starts with tenants

Property management is a volume business. Every rented unit produces a steady stream of small jobs: a dripping tap, a question about the heating bill, a lost key, a notice to end the lease. None of them is hard. Together they fill the day of a property manager, and most of the effort goes into reading, sorting and forwarding.

In Switzerland the stream is especially large. 58.0% of the Swiss population lives in a rented home, against 31.6% across the EU, according to Eurostat. The Swiss Federal Statistical Office counts 2.4 million households, 61% of the total, in a rented or cooperative flat at the end of 2024 (BFS). Germany sits close behind at 52.8%. Spain is the opposite case at 26.3%, but there the work moves to the property administrators who run the owners' associations of every apartment building: owner requests, community fees, contractors and annual accounts.

The tools are the same in all three markets: a shared inbox, PDFs, a property management system and a lot of spreadsheets. An agent does not need a new platform. It needs read and write access to the systems your team already uses, and clear rules for what it may do on its own. Our real-estate page lists the systems we build for property managers, agencies and real-estate funds. This article goes one level deeper, process by process.

Part 02

Process 1: tenant and owner requests

A property manager with a few thousand units receives requests through a shared inbox, a web form, phone notes and messaging apps. For each one somebody finds the building and the unit, decides how urgent it is, calls the right contractor and answers the tenant. Most of that time goes into sorting, and sorting is where an agent is strongest.

The agent reads every new request and matches the sender to a tenant and a unit in your property management system. It classifies the request as a repair, a question about rent or costs, a complaint, a notice or an owner matter. It sets the urgency: a burst pipe, a heating failure in January or a tenant locked out comes before a squeaky door. For a repair it opens a job for the contractor you normally use for that building and trade, within the amount you allow. Then it drafts the reply in the tenant's language and logs everything on the unit's record.

What stays with a person: approving replies until a category has proven itself, anything about rent reductions, disputes, terminations or neighbour conflicts, any job above the amount you set, and the choice of contractor when the usual one cannot come.

What the agent needs: read access to the inbox, the tenant and unit register from the property management system, the list of contractors per building and trade with their approval limits, and a few hundred past requests together with how your team handled them. Those past cases become the evaluation set, the yardstick that shows whether the agent sorts as well as your best property manager.

What can go wrong: a tenant writes from a partner's address and the agent matches the wrong unit. Water damage described casually ("there is a stain on the ceiling") gets low urgency. A reply promises a visit date the contractor never confirmed. The fixes are plain. Anything below the confidence threshold goes to a person, reply drafts never commit to dates or costs the system does not hold, and someone reviews a sample of handled requests every day in the first months.

Let the agent sort, draft and route. A person decides anything that costs money or touches a tenant's rights.

Alpgency

Part 03

Four more processes worth automating

Process 2: lease and document processing

Leases, addenda, handover reports and invoices arrive as PDFs or scans. The agent extracts the parties, the rent, the deposit, the indexation clause, notice periods and key dates, compares them with the property management system and flags every difference. It sets alerts for lease ends, rent reviews and deposit returns, and it lets your team search all documents in plain language.

A person interprets non-standard clauses and approves any change to the master record. The agent needs the documents, the system records, the list of fields that matter to you and around fifty annotated examples. It goes wrong on handwritten handover reports, old leases with amendments in the margin, and two versions of the same contract. So every extracted field links back to the page it came from, and low-confidence fields go to review.

Process 3: listings and lead follow-up

The agent drafts listing texts from the unit data, answers every portal enquiry on the same day with availability, the documents an applicant needs and open viewing slots, books viewings into the calendar and follows up afterwards. Each contact lands in the CRM with its history.

Tenant selection stays with a person, as do price questions and the final check of legal statements in the listing, such as floor area. The agent needs the unit data, the calendar, the enquiry feed or inbox and your templates. Typical failures: a listing mentions a balcony the flat does not have, two viewings land in the same slot, a follow-up reaches someone who asked not to be contacted. Applicants also send salary slips and, in Switzerland, extracts from the debt enforcement register. That is sensitive data, and the agent should file it, never score people on its own.

Process 4: service-charge statements and invoice checks

Every contractor and utility invoice has to land on the right building, cost type and period. The agent matches each one, checks it against the order or contract and against previous years, and flags duplicates, wrong buildings and unusual price jumps. At year end it prepares the allocation per unit with your keys (floor area, meter readings), drafts the statement and the cover letter, and later answers tenant questions with references to the underlying invoices.

A person signs every statement and decides what may be charged to tenants under the lease and the law. The agent needs the invoices, the chart of accounts, allocation keys, meter readings, the cost clauses of each lease and last year's statements. The risks: a cost type charged that the lease does not allow, meter readings swapped between two units, an invoice booked to the wrong year. These errors reach tenants as money and can be contested, so here the agent checks and drafts while a person signs.

Process 5: portfolio reporting for owners and funds

The agent pulls occupancy, arrears, the rent roll, costs and open tickets per property from the property management and accounting systems, builds the owner report or the fund report on schedule and writes a short comment on each change ("two more vacant units in building C after the renovation"). It reconciles the totals with the source before anything goes out.

A person approves the commentary and every number that reaches investors. The agent needs access to the source systems, the report template your owners know and one sheet that defines each figure. It goes wrong when two systems define vacancy differently, when an export is stale, or when the comment sounds certain where the data is thin. When a number does not reconcile, the report goes out with the alert instead of going out wrong.

Part 04

Which process to automate first

Two questions settle the order. How often does the work come in, and what does one error cost? High volume with a cheap error is where an agent earns its keep from the first month, because a person still reviews the drafts and the time saved is large. High risk with lower volume comes later, and starts with checks instead of output.

Part 05

Tenant data: GDPR and the Swiss revFADP

Tenant files hold names, contact details, payment history, sometimes health information (an accessibility need, a doctor's note about mould) and debt records. All of it is personal data, and an agent that reads it falls under the same rules as your staff and your software.

In the EU, and so in Spain and Germany, the GDPR applies. If a provider runs the agent for you, that provider is a processor. The EU's Your Europe portal spells out what follows: a written contract with sufficient guarantees, a data protection impact assessment where processing is high risk (automated decision-making is one of the named cases), and a breach report to the authority within 72 hours.

In Switzerland the revised Federal Act on Data Protection has applied since 1 September 2023 (FADP on Fedlex). Four articles matter most for an agent:

• Article 9: you may hand processing to a processor only if it processes the data as you would, and you must satisfy yourself that it can guarantee data security. Sub-processors need your prior approval.
• Article 16: personal data may go abroad only to a country with adequate protection, or with safeguards such as approved contract clauses.
• Article 21: a decision based exclusively on automated processing, with a legal consequence or a considerable adverse effect on the tenant, must be disclosed, and the tenant can ask for a person to review it.
• Article 24: a breach likely to cause high risk must be reported to the Federal Data Protection and Information Commissioner as quickly as possible.

Swiss fines target people. Under Articles 60 to 63, a fine of up to CHF 250,000 can be imposed on the responsible private person, for example for wilfully handing data to a processor without meeting Article 9. And if tenants chat with an AI system in the EU, Article 50 of the EU AI Act requires telling them so from 2 August 2026.

Part 06

Where to start

Our clients include Im Lot Immobilien, a Swiss property company. Every engagement starts the same way: one process, one measurable baseline, one price agreed before the work begins.

1. Count the work.Requests per week, minutes per request, the share that ends with a contractor job. Without a baseline nobody can tell whether the agent helps.
2. Collect real cases.A few hundred past requests, leases or invoices, with what your team did with each. They become the test set.
3. Free Audit.We map the process, the data and your tools, rank the use cases and give the first one a fixed price. The findings are yours to keep.
4. Demo, then build.A working demo on your real data, then the build to production. One fixed price, agreed after the Audit.
5. Run it your way.After go-live you choose: a full handover to your team, a fixed monthly support fee, or us as your tech partner on a fixed schedule.

FAQ

Frequently asked questions