The 7:43am leak
It always arrived in the same shape. A tenant messaging at 7:43am, before the office was open, with two blurry photos and one line: “Water coming through the kitchen ceiling, please help.” For eight years I was the person who picked that up. And the first thing I did, every single time, was guess.
Was it the bath overflowing in the flat upstairs, or a joint that had finally let go? Was the tenant worried about a drip, or standing in an inch of water with the light fitting crackling? You could not tell from the message. So you rang the tenant, who did not answer because they had already left for work. You rang the plumber, who was on another job and would “try to swing by later”. You wrote a note to chase it, and then five more jobs landed on top, and by four o’clock the note was gone. Two days later the landlord rang to ask why nobody had told him his flat had a hole in the ceiling.
That is the manual version of maintenance in a letting agency. Triage by instinct, three phone calls, and a follow-up that survives only if you happen to remember it. Nobody is bad at the job. There is simply too much of it, arriving in the wrong order, on the days you can least afford to stop and think. Repairs are the loudest inbox a property team owns, and they are the one nobody has time to design properly.
What AI maintenance triage actually does
AI maintenance triage for a property management team is not one clever button. It is a short, boring, reliable pipeline that does the first twenty minutes of work you used to do by hand, before a person even opens the message. The point is not to sound futuristic. The point is that the leak at 7:43am gets read, understood and moved forward while you are still on the train.
Here is what it reads. The tenant’s message, in whatever channel it landed. The two photos, which it can actually look at rather than skim. The property record: who the landlord is, what the tenancy says about repairs, whether there is a managed budget and a spend threshold, and which contractor covers that postcode and that trade. The history, so it knows this is the third time this shower has been reported since March.
And here is what it writes. A draft works order with the fault described in plain terms, the access details, and the photos attached. A suggested contractor from your own approved list, matched to trade and area, not a random search. A draft reply to the tenant that sets expectations honestly. A short note for the landlord’s file. None of it sends on its own for anything that carries risk. It stages the work so a person approves and presses go, instead of starting from a blank screen and a bad memory.
The saving is not really the typing, though the typing adds up. The saving is that nothing sits unread in a queue because the office happened to be busy that morning. The first pass is done. What reaches a person is already sorted, already briefed, and already labelled with how urgent it thinks the job is.
Emergency or annoying? The classification that matters
The whole thing lives or dies on one decision: how bad is this, actually? Get it wrong in one direction and you send a plumber out at weekend rates for a dripping tap. Get it wrong in the other and a genuine escape of water sits in a routine queue for two days while a ceiling comes down.
A useful classification has four buckets. Genuine emergency: no heat or hot water in winter, an active leak, an electrical fault, anything that threatens the property or the tenant’s safety. Urgent: things that are not dangerous today but will get worse and dearer if left, like a failing boiler in October. Routine: the cupboard door, the slow drain, the blind that will not sit straight. And tenant-fixable with guidance: the tripped switch, the boiler that needs repressurising, the smoke alarm chirping for a battery. That last bucket quietly saves the most money, because a good first reply solves the problem before a contractor is ever booked.
This is where photos earn their place. A description says “a bit of water”. The picture shows the ceiling already sagging, or a fuse board with a scorch mark, and the classification moves up a bucket where it belongs. It works the other way too. What sounded like a flood in a panicked 7:43am message turns out to be a manageable drip under a sink, and the tenant gets a calm reply instead of a callout they did not need.
The hard rule sits underneath all of it: anything ambiguous rings a person. If the system is not confident, it does not quietly pick a bucket and hope. It escalates. One ambiguous report is all it takes to justify keeping a human on the gate, and that is a feature, not a failure. Repairs also sit under real legal duties. Landlords carry repair and habitability obligations, and in the Renters’ Rights Act 2025 era the paper trail behind a repair decision matters more than it ever did, because a tenancy is harder to end and easier to scrutinise. A clean, timestamped record of what was reported, when, and what you did about it is worth having long before anyone asks for it.
Chasing: the part everyone hates and nobody budgets
Reporting a repair is the easy bit. Chasing it to done is where the quiet hours of a property manager’s week disappear, and it is the part no proposal ever costs in. The contractor said Tuesday and did not come. The tenant was not in. The invoice arrived with no note about what was actually fixed. The landlord was never told it was closed. Multiply that by a portfolio and you have a full-time job made entirely of loose ends.
Automated chasing handles the follow-up you never have time for. It nudges the contractor when a promised date slips. It asks the tenant, after the visit, whether the job was actually done, and it collects the completion evidence: the photo of the finished work, the note on what was replaced. It matches the invoice against the works order that authorised it, so an unexpected line does not get paid on autopilot. Then it closes the loop with the landlord in a short, plain update rather than a fortnight of silence.
The by-product is the thing you would never build on purpose but always wish you had. Every step is logged, so the audit trail is not a separate task somebody has to write up. It is simply what falls out of doing the work in the open. When a dispute lands six months later, over a deposit deduction or a repair that supposedly never happened, the answer is already assembled instead of reconstructed from memory and a shoebox of emails.
Keeping the human on the gate
The temptation, once this works, is to automate everything. That is the mistake. Some decisions have to stay with a person, and being clear about which ones is what makes the rest trustworthy.
We refuse to automate three things outright. Emergency judgment, because the cost of a machine deciding a sparking fuse board is “routine” is a person’s safety, not a bad review. Spend authority, because money leaving a landlord’s account should always cross a human desk above a set threshold. And anything tribunal-facing, because a decision that might be examined by a court or the First-tier Tribunal deserves a name attached to it, not an algorithm’s shrug. Data discipline matters here too. Tenant messages and photos are personal data, and the ICO’s AI and data protection toolkit is a sensible place to start on what gets sent where, who can see it, and how long it is kept.
If you want to try this, start narrow. Turn it loose on routine repairs only, the cupboard doors and the slow drains, where a wrong guess costs a callout and not a catastrophe. Leave every emergency ringing a person from day one. Measure it for a fortnight: how fast reports were acknowledged, how many contractor chases the office no longer had to make, how many jobs closed with evidence on file. If the saved time is real, widen the buckets. If it is not, fix the process before you add more software to it.
This is the property management automation work we do, built into the systems you already run, your inbox, your phones, your management software, rather than another platform login to remember. The founder ran UK lettings for eight years before starting this, which mostly means I know exactly which follow-up gets forgotten first, because for a long time I was the one forgetting it. If maintenance is the loudest part of your week, that is usually the right place to start.
