AI in property management: what AI can do in operations today, and what it should not do alone

AI in property management is software that reads a property's own data, such as leases, bank transactions, tenant requests and building registers, and proposes or carries out the next step in operations. The value is in repetitive work: tenant requests, bank reconciliation, lease checks and month-end. What separates useful AI from a black box is three things: visible sources, human approval of anything binding, and a log of everything that happened.

Task	What AI does	What a person does	Control
Tenant requests	Sorts them, answers routine questions, opens the case on the right unit	Handles unclear cases, approves replies with consequences	Low risk, can run alone within set limits
Bank reconciliation	Proposes a match between payment and unit and explains why	Resolves items that do not match	Proposal, a person closes
Lease review	Reads leases and flags terms that need a look	Assesses and decides	AI flags, a person assesses
Rent roll and indexation	Checks leases against the rent roll and finds, for example, unused index-linked adjustments	Decides whether to adjust	Proposal with source
Maintenance	Opens the case, contacts the contractor, collects photos and quotes	Approves larger costs	Spending limits
Month-end and reporting	Collects receipts, finds blocking items, drafts the report	Approves closing and report	Draft, a person approves
Letters with legal effect	Prepares drafts with basis and deadline	Always approves before anything is sent	Always human approval

From AI tool to agentic property operations

AI enters property operations at three levels. They look alike in demos but do very different things day to day.

1. AI features in an existing system. AI helps with one step, such as a suggested account code or budget. The employee still drives the case.
2. AI assistant. AI drafts replies and letters, and an employee approves and sends them. It saves writing time, but the work does not move.
3. AI agent. AI carries the case from start to finish: opens it, finds the basis, contacts the right people, prepares the next step and asks for approval when something is binding.

We call the third level agentic property operations. Read the full definition in What is agentic property operations?, and see the difference between the three models in AI agent vs. traditional property management software.

Five places where AI makes a difference today

1. Portfolio truth: companies, registers and properties

When you take over a portfolio, the first question is: what do we own, and which units belong to which companies? AI can link company registers, addresses, property IDs, building data and internal notes. The value is not a nice map. It is flagging uncertainty: what is confirmed, what is likely, and what needs a manual check.

2. Rent roll and leases

If area, rent, indexation or deposit is wrong in the rent roll, the error spreads to budgets and reports. AI can check leases against the rent roll and find missing adjustments, empty fields and numbers that do not match the documents. It should not fix things silently. It should show the deviation, the source and the recommended action.

3. Month-end and reporting

AI can collect bank transactions, invoices, rent charges and reconciliations and highlight the items that block closing. Instead of hunting for the error, the team starts with a prioritised list: missing receipts, unmatched payments, unexpected deviations and items without approval.

4. Maintenance and contractors

An agent can take a fault report, open the case, contact the contractor, follow up and collect photos and quotes. The goal is not for AI to choose the contractor alone. The goal is a clear view of what has happened, what is urgent and what affects the budget.

5. Tenant communication with a trail

AI can answer routine questions, send status updates and draft replies. In professional management, communication must be correct and traceable, so the agent must show which information it used and which channel the message went through, such as an app, email or Denmark's Digital Post.

Five rules for AI with control

1. Sources must be visible. Every proposal must trace back to a lease, bank transaction, register or legal provision.
2. Binding actions require approval. A person says yes before anything with legal or financial consequences leaves the system.
3. Everything is logged. Who did what, when and on what basis.
4. Proposals can be rejected and corrected. And the corrections are used to improve.
5. Facts and assumptions are kept apart. A report must show what is confirmed and what is an estimate.

What a good answer from an agent looks like in practice (example):

• 43 of 46 units match the lease and rent roll.
• 2 units are missing documentation for the latest adjustment.
• 1 payment matches on amount but not on reference.
• 3 maintenance cases should go into the next owner update.

What AI should not do alone

• Send terminations, rent increases or formal notices without approval
• Approve larger costs or payments
• Settle disputes with tenants
• Decide who gets a unit without human judgement

Here AI should prepare, so the responsible person can decide faster. The final decision stays with the landlord or property manager.

What do the rules say?

When AI processes tenant data, GDPR applies, and depending on the use, the EU AI Act may apply too. The European Commission's timeline shows that the Act's transparency rules (Article 50) apply from 2 August 2026, and the rules for high-risk systems in Annex III from 2 December 2027 (AI Act Service Desk). We cover roles, requirements and responsibility in AI agents in property management: control, GDPR, the EU AI Act and responsibility.

How to get started

1. Pick one portfolio with both system data and documents.
2. Get the basics right: companies, properties, units, tenants and responsibilities.
3. Pick one process to start with, such as tenant requests, bank reconciliation or lease review.
4. Require the AI to show sources and uncertainty for every proposal.
5. Keep human approval on finance, legal and communication until the patterns are stable.
6. Measure before and after: response time, unresolved items at month-end and the share of proposals approved without changes.

How Matrikel uses AI

Matrikel is built as one agent for all of property operations, currently for Danish rental property. This is what we say on matrikel.io:

• The agent lives in your Messages. Matrikel messages you when something needs your attention, answers tenants around the clock and gets hold of your contractors. You approve sensitive steps with Face ID (matrikel.io/en).
• Nothing binding without your yes. Matrikel never sends a reminder, terminates a tenancy or closes the books without your approval (matrikel.io/sikkerhed, Danish).
• Every action is logged. Timestamped, explainable and available afterwards, including to your auditor (matrikel.io/sikkerhed, Danish).
• The bank is read, not controlled. Bank data is read through a licensed PSD2 provider, and Matrikel cannot move money (matrikel.io/en/finance).
• All channels in one place. Tenant app, portal, email, phone, and Digital Post and e-Boks (matrikel.io/en/pricing).
• Price. DKK 89 per unit per month excluding VAT, bookkeeping included (matrikel.io/en/pricing).

Want to see it on your own numbers? Book a demo.