AI for property management companies: use cases and how to start

What does AI actually do?

Property management is full of repeated work. Tenants ask the same questions, repairs arrive by phone, email and messaging apps, and leases and certificates pile up in folders. AI is good at reading, sorting and drafting this material, and the strongest use cases start with the tasks your team repeats every day. Pick those first. Each one is small, easy to measure and low risk, and together they free up hours every week for tenant relationships, owner service and the decisions that need a person.

In practice, an AI solution sits between your people and your systems. It reads incoming messages and documents, checks them against your records, then drafts a reply or a task for someone to approve.

Assistant     MCP server       CMS
    |              |             |
    |--tools/list->|             |
    |<--tools------|             |
    | name, role,  |             |
    | parameters   |             |
    |              |             |
    |--tools/call->|             |
    | chosen tool  |---write---->|
    | + content    |   draft     |
    |<--result-----|<------------|

Here is how a repair request moves through such a solution. A tenant messages that the boiler has stopped working. The assistant reads the message, finds the tenancy and property in the PMS, checks whether the issue is urgent and asks for any missing detail, such as photos or access times. It then drafts a work order for a contractor from your approved list and sends it to a member of staff. The team approves it, and the tenant gets a clear update. The assistant only touches the records its account is allowed to reach.

Most property software can already be connected to other tools. Lettings platforms, accounting packages and maintenance apps usually publish an official integration, so that a custom solution can read and update records without anyone copying data by hand. Each connection runs through a restricted account that you control. It sees only the records its job needs, every change is logged, and access can be switched off at any moment without affecting your core systems.

How can AI handle tenant communication?

Tenant communication is where most teams feel the pressure first. Questions arrive all day: when is rent due, can I book a viewing, what is happening with my repair? An assistant can answer these from your own records, in your tone, at any hour. It can also sort the inbox, summarise long threads and draft replies for your team to check. The gain shows most on repetitive work, where one person used to spend the morning answering the same few questions.

A solution only works if it fits how your team really handles messages. That is why we begin with the Embedded Week , in which we sit with your team for a week, watch how enquiries, repairs and hand-offs really flow, and map where the time goes. Then we build the solution for your company and train your staff to run it.

The work is shared sensibly between the AI and your team. The assistant drafts, sorts and reminds, while decisions about tenants, exceptions and anything sensitive remain with your people.

How does AI triage maintenance requests and contractors?

Maintenance is a strong first use case because the steps are clear. An assistant reads each request, classifies the issue, judges urgency from your own rules and gathers the missing details. It can then propose a contractor from your approved list and prepare the work order. Each rule, such as which faults are emergencies or what spending needs sign-off, is yours to set, and the assistant follows it every time.

Consistency is the second benefit. Different team members triage differently, and busy periods change the answer. A solution applies the same checks to every request, records why it chose what it did and shows its working to the person approving. It also keeps tenants informed at each step, which cuts the chasing calls. We explain how agents differ from simple automation and chatbots in a separate guide.

Generic tools only go so far, though. A standard maintenance app does not know your contractors, your owners' limits or how your team handles an out-of-hours call. The smoothest results come when the solution is built around those details, rather than asking your staff to adapt to the software.

So when we build for a property company, the assistant writes in your tone, follows your approval limits and hands over to a named person whenever a case is unclear. The guidance for each task sits beside it, so the assistant reads your rules before it acts, and your team can change them.

{
  "name": "saveArticle",
  "description": "Saves an article as a draft, in French and native English.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "fr": { "description": "Title, standfirst, Markdown body, FAQ" },
      "en": { "description": "Native English text, not a translation" }
    }
  }
}

The instructions therefore work as a rulebook. They remind the assistant that a repair over the limit needs approval, and that a tenant should always be told what happens next. Connecting software to an AI gives it access to your data, while these rules teach it to work the way your company works. The same applies to lettings, sales and block management alike.

Can AI read leases and chase arrears?

Leases and compliance documents are a natural fit because they are long, similar and full of dates. AI can read a tenancy agreement and pull out the term, rent, break clauses and notice periods. It can check that certificates, such as gas and electrical safety records, are on file and flag those nearing expiry. It can summarise a long inspection report for the property manager in a few lines. Staff still read anything that matters, but they start from a summary rather than a blank page.

Arrears work follows the same pattern. The assistant spots late payments in the ledger, drafts a polite first reminder in your tone and prepares a follow-up schedule. Escalation, payment plans and anything sensitive stay with a person.

Owner reporting is the third piece. Owners want to know what happened to their property: rent collected, repairs completed, upcoming works. Today someone assembles that from several systems each month. A solution can pull the figures, draft a plain-language update and attach the supporting documents, ready for the manager to review and send. It does the assembly and the writing, and your team keeps the relationship and the explanation.

What risks come with AI in property management?

Start with accuracy. AI can write a confident reply that is wrong, so anything that goes to a tenant, owner or contractor for the first time should be reviewed by a person. As trust builds, low-risk replies, such as confirming a viewing time, can be sent automatically while the rest stays under review. Deciding where that line sits is part of the design.

Confidentiality is the second limit. Tenancy files hold names, contact details and sometimes sensitive circumstances. Limit what the AI can see to what the task needs, and check how any provider stores and uses your data. Rules on personal data differ by jurisdiction, so check yours.

Keep a person in charge of judgement. AI supports professional judgement and does not replace it. Decisions on evictions, deposits, disputes and safety belong to qualified staff. Nothing here is legal advice, and rules for tenancies differ between countries and cities.

Finally, keep an audit trail. The solution should log what it read, what it drafted and who approved it. If something goes wrong you can see why, and an owner or tenant can be given a clear account. Start with a few workflows, review them weekly and widen the scope once they are stable.

Book a meeting to see where AI fits

Setting up the first connection takes little development. Most of the effort goes into deciding what the assistant may do, then writing your house rules down clearly.

In the Embedded Week we find the rules your team follows without writing them down, then build them into a solution made for your company and train your staff to run it.