How to automate arrears chasing in property without getting it wrong

The short answer. Automate the early, boring end of arrears chasing: the reminder that goes out on day three, the drafted follow-up on day ten, the file note, the escalation flag. Keep a human on the tone, on the decision to escalate, and on anything moving towards legal action. Chasing money owed is regulated territory in the UK, and an automation that keeps sending after a person should have picked up the phone is worse for you than the arrears were.

I spent years in commercial property, then a decade building and automating businesses, and arrears chasing is one of the clearest wins I see in property management. It is also one of the easiest to get wrong.

Why arrears chasing suits automation

Because most of the work is not the difficult conversation. It is remembering.

Sit with a property manager for a morning and the pattern is obvious. Little of the arrears day goes on judgement. It goes on noticing that an account has slipped, finding out whether anyone has already chased it, working out what was said last time, writing a message close to the one written last week for someone else, and logging what happened so the next person is not starting cold. That is the job. All of it is repetitive, none of it is skilled, and all of it gets dropped first when the week gets busy.

The awkward conversation, meanwhile, is a small slice of the total hours and the only part that needs you. Which is exactly the wrong way round from how most desks are run.

What to automate, in order

Work outwards from the safest end.

1. The noticing. A daily check of the ledger that flags accounts that have moved into arrears, with the amount, how long, the history and whether a payment plan is in place. This one has no downside. Nothing goes out to anyone, so it is impossible to embarrass yourself with it.

2. The record. Every contact logged automatically: what went out, when, to whom, what came back. Most firms do this badly, and it is the thing you will wish you had if a matter ever goes anywhere formal. Automate the logging before you automate the sending.

3. The first reminder. A short, polite, factual message a few days after the due date. Amount, period, how to pay, who to talk to if there is a problem. Most arrears at this stage are an oversight, a failed standing order or a change of bank. Treating that as a debt problem damages a decent relationship for no reason.

4. The drafted follow-up. Not the sent follow-up. AI drafts the next message from the account history and the lease, a person reads it in ten seconds and sends it. You get most of the time saving with none of the risk, and the drafts get better as you correct them.

5. The escalation flag, not the escalation. The system tells a named person that an account has hit the threshold. The person decides what happens. Never let a workflow serve a notice or instruct a solicitor because a date arrived.

Where automation has to stop

At the point where the message stops being a reminder and starts being pressure.

This is the part the software vendors skip entirely, and it is the part that matters. In the UK, how you chase money owed is constrained by law. There are limits on the frequency and manner of demands, and there are steps expected of you before a claim is issued. Some of those constraints depend on whether you are chasing a residential occupier or a business tenant, and the residential rules in particular have been through significant change. I am not going to state the current position for you, because it moves and because getting it wrong is expensive. Have your solicitor read your escalation sequence before you switch it on, and have them read it again after any change in the law that affects possession or debt recovery.

What I am comfortable telling you is the shape of a safe process:

Cadence over volume. An automation that sends daily because sending is free is the single most dangerous thing you build. Frequency itself is part of what makes chasing improper. Space it out.
One channel at a time. Email, then a letter, then a call. Firing email, SMS and a letter simultaneously reads as pressure even when each message is polite.
A working off switch. The moment a tenant replies, disputes the sum, mentions hardship or agrees a plan, automated chasing stops and a person owns the account. If your build lacks that switch, it is not finished.
No automated threats. Nothing generated by a machine should reference court action, possession, or a solicitor. Those go out over a named person's signature, after that person has thought about it.
Nothing about the person. Automated content states the amount, the period and how to pay. It does not characterise the tenant or the reason.

The tone problem nobody writes about

Arrears messages fail because they are written in the wrong register, and AI amplifies whatever register you give it.

Ask a model to write a rent arrears letter with no direction and you get something faintly American, faintly legalistic and colder than anything you would have written yourself. Send that at scale and you turn a hundred solvable oversights into a hundred defensive tenants. Then the calls come to your team instead.

The fix is briefing rather than prompting. Give the model three or four of your own best chasing emails, the ones that got a payment without a row, and tell it that is the standard. Tell it the reader is a person who has most likely made a mistake rather than a person avoiding you. Tell it to stay under a hundred and twenty words, state the facts and offer a route to talk. Then read the first thirty drafts before you trust the thirty-first.

What if you do not have the software?

You get most of the benefit with a spreadsheet, a shared inbox and an hour of setup.

Vendor content on this topic assumes you already own their platform, which is not much help to a firm running on a property management system that predates all of this. You do not need a new platform to start:

Export the arrears list on a schedule and have AI turn it into a prioritised worklist with a suggested action per account.
Keep a document of your approved message templates by stage, and have AI fill them from the account details rather than write from scratch.
Have every sent message copied to one address, and let AI summarise the week's chasing activity into something a director reads in two minutes.
Ask for a weekly list of accounts where nothing has happened for fourteen days. That single report finds more money than most software does.

Do that for a quarter, learn where the process leaks, and then buy software knowing precisely what you need it to do. Firms that buy first spend a year configuring a system around a process they never examined.

Frequently asked questions

Does automated chasing recover more money? The early reminders do, because a large share of arrears is administrative rather than deliberate, and speed alone fixes those. Deliberate non-payment is not moved by more messages. It is moved by a person, a conversation and a plan, which is why the aim of automation here is to hand your team more hours for that conversation.

Should AI write the letter that goes to a tenant in arrears? It writes the draft. A person sends it. That split gives you the time saving and keeps a human accountable for tone and accuracy, which is where the risk sits.

Is it safe to put tenant data into an AI tool? Only with a tool your firm has approved, on a business tier, with a clear answer on what happens to what you type. Arrears data is sensitive: it is financial information about identifiable people, often in difficulty. Consumer free tiers are the wrong place for it.

What about automated calls and texts? Treat them as higher risk than email and take advice before using them. They land more intrusively, the rules around automated contact are stricter, and the reputational cost of getting the cadence wrong is higher.

How do I stop the automation running over someone in genuine hardship? Build the stop switch first, before the sending. Any reply, any dispute, any mention of difficulty pulls the account out of the sequence and onto a named person's list. Test that path before you go live.

What is the first thing to build? The daily flag and the automatic file note. Neither sends anything to anyone, both save real time, and together they give you the clean record everything else depends on.

Arrears is usually the first thing I build with a property management team, because the hours are visible and the win is quick. If you want your team trained on this and the sequence built properly, book a session.