A tenant's boiler dies at 7:40 on a Saturday evening. Nobody at the management office will read that message until Monday at 9. By then the flat has been cold for two days, the tenant has called the emergency line twice, and the review is already written. A property management chatbot exists to close exactly that gap, and to leave everything else alone.
Written by the ChatterMate team. We build an open-source AI support agent, so we have a stake here. We will say where a bot should not be trusted, because that is where most of the damage happens.
Why property management is an after-hours business
People notice what is broken in their home when they are in it. That means evenings and weekends.
Two UK sources put numbers on it. Lettie, a property management platform, analysed a summer sample of maintenance cases and found that 50.0% were first logged outside Monday to Friday, 9am to 5pm, and 17.4% at weekends. For heating and hot water the figure was 73.1% outside office hours. Property118 reported it in September 2026, without a stated sample size.
Adiuvo, an out-of-hours provider, says 81.3% of tenant calls land outside standard office hours; of nearly 226,000 property-related calls it handled in 2024, 183,752 came after hours. It also reports that 57% of landlords it surveyed have no after-hours support at all. Both firms sell services built around this problem, and neither publishes its method. Treat the exact percentages as directional. The direction is not subtle, though: half or more of the demand arrives when your office is shut.
And the same Lettie data says more than one in eight cases was a recurring or previous problem. Repeat issues are where tenants stop being patient.
What a property management chatbot should actually do
Not much. That is the point. A chatbot for a letting agency or a block manager earns its place on four jobs, and we would resist adding a fifth.
1. Take the maintenance report properly
"The tap is broken" is not a work order. A contractor needs the room, the symptom, whether water is actually escaping, whether there is a shutoff, and a photo if the tenant can send one. A bot asks those questions at 11pm without getting tired, and hands your team a complete ticket instead of a three-word email. This is the unglamorous win. It saves the second phone call on Monday.
2. Answer the questions that already have written answers
Bin day. How to bleed a radiator. Whether pets are allowed. When rent is due and how to pay. What the notice period is. Your tenancy pack, house rules and move-in guide already answer these. A bot grounded in those documents answers from them and shows which page it used. If the document does not say, the bot says it does not know and passes the question on. We wrote about training a chatbot on your own docs separately; the property version of it is mostly about getting old PDFs out of people's inboxes and into one source.
3. Sort urgent from not urgent
This is the job that matters and the one most easily done badly. We will come back to it below.
4. Hand over to a human with the context attached
When the bot cannot help, the tenant should not have to repeat themselves. The person picking it up sees the whole conversation. We covered the mechanics in our post on human handoff that does not frustrate customers. In property it is not a nicety. A tenant describing a leak for the third time is a tenant who is about to complain to the landlord.
The urgent-versus-routine line is the whole product
Every property business has a list of things that cannot wait: no heating in winter, water escaping, a gas smell, no power to a flat, a lockout, a broken external door lock, a smoke alarm sounding. Your own list will differ by country, building and contract. Write it down. Not in a head, on a page.
Then decide what the bot does when a message touches that list. Our view, which you can argue with: the bot should never reassure and never triage safety itself. If the tenant says gas, the bot gives your emergency instruction and the number to call, in the first sentence, and pages a human. It does not ask five follow-up questions first. It does not say "a technician will be there shortly" unless a person has actually committed to that.
The failure we worry about most is not a wrong answer about bin day. It is a polite, confident reply to a message that needed a phone call. So test for it. Build twenty or thirty messages that should trigger the urgent path, including the badly worded ones ("theres water coming thru the ceiling lol", "smells weird in the kitchen"), and run them before go-live. Our chatbot testing guide walks through how we set that kind of test up. Re-run it every time you change the documents or the instructions.
One more habit: tell tenants plainly that they are talking to an AI, and that for anything dangerous they should call the emergency number, which you print on the first message. That costs nothing and it is the right thing to do.
Where a bot should stay out
Some property conversations are not support questions.
Anything about applicants or who gets a home. In the US, HUD published guidance in May 2024 on how the Fair Housing Act applies to tenant screening and to housing advertising that uses algorithms and AI, as HousingWire summarised at the time. The reporting we read does not discuss chatbots specifically, so we are not going to claim it does. Our position is simpler than any legal reading: a support bot answers questions about a property and a process. It does not rate applicants, suggest who would suit a building, or steer anyone toward or away from an area. If your market has fair housing or equality rules, which most do, get advice from someone qualified, because we are not lawyers.
Disputes. Deposit deductions, rent arrears, noise complaints between neighbours, repair responsibility arguments. The bot can explain the process and book a call. It should not take a side or promise an outcome.
Anything involving a vulnerable tenant. If a message suggests someone is unwell, frightened or in distress, the answer is a person, fast.
Entry, keys and access. Telling a contractor or a stranger how to get in is a security decision. The bot can log a preferred time. Your team confirms access.
Owners and landlords are a second audience
Block managers and agents also answer to owners. "Has the roof quote come in?" "What is the service charge for this year?" "Can I see the last inspection report?" The same grounding approach works if you keep a separate document set for owners and a separate bot for them, so a tenant never gets an answer drawn from a landlord document. Keep those two knowledge sources apart on purpose. Mixing them is how private information ends up in the wrong chat.
For a small agency this is also where self-hosting earns attention. Tenant messages contain names, addresses, photos of the inside of people's homes and details of when they are out. Some agencies prefer that data stays on infrastructure they control. ChatterMate is open source and can be self-hosted, which is one reason we think it fits here. It is not the only reason to pick a tool, and plenty of agencies will be fine with a hosted one that has a clear data retention policy. Ask any vendor, us included, how long conversations are stored and how to delete them.
A rollout that will not embarrass you
The worst way to start is switching the bot on for every tenant on a Friday. Here is the order we would use instead.
- Export the documents tenants actually ask about: tenancy pack, house rules, appliance guides, move-in and move-out checklists, emergency instructions. Delete the outdated ones first.
- Write the urgent list, with the exact words the bot should use and the number it should give.
- Pull the last 100 real tenant messages from your inbox. Run them through the bot in a test environment. Mark every answer right, wrong or "should have gone to a person".
- Go live on one building or one portfolio. Watch the transcripts daily for two weeks.
- Only then widen it.
If it feels slow, remember that a single missed heating call in January costs more than the whole setup week.
Measure two numbers and ignore the rest at first. How many conversations ended with a complete maintenance ticket and no human typing. And how many urgent messages reached a human within the time you promised. The second one matters more. We cover the first in our post on measuring deflection and ROI, and the overnight angle in honest after-hours support.
What we have not built, and will not pretend to
We are not a property management system. ChatterMate does not replace AppFolio, Buildium, Rentman or whatever holds your ledger and work orders, and we are not claiming a ready-made integration with any of them. What it does is answer from your documents with citations, hold the conversation, and hand it to a person. Getting a finished ticket into your maintenance system today means a handoff your team completes, or a connection you build through our integrations and API. If you already live in a property system with its own tenant bot, compare honestly before adding another tool.
We would rather say that now than have you find it in week three.
Try it on one building
Pick a single property, load its tenancy pack and house rules, write the urgent list, and test it against real messages before any tenant sees it. ChatterMate is open source, so you can read the code, and the free plan needs no card, which makes it a cheap way to find out whether your documents are good enough. Start at chattermate.chat, or see how other trades handle the same problem in our industry guides, such as appliance repair and hotels.



