AI Debtor Chasing
The follow-up that keeps not happening, happening — politely, consistently and on time. Any reply stops the sequence and hands the conversation to a person. No automated threats, ever.
Almost Nobody Refuses to Pay. They Just Are Not Asked.
Late payment is overwhelmingly an administrative failure on both sides — and the creditor’s side of it is entirely fixable.
The mental model most people carry is adversarial: a debtor who will not pay and a creditor who must apply pressure. That describes a small minority of cases. The overwhelming majority of overdue invoices are overdue because something administrative went wrong — it went to the wrong address, it is sitting in an approval queue behind someone on leave, it arrived during a busy week and got buried, the payment run is fortnightly and it missed the cut. None of that needs pressure. It needs a reminder.
But the reminder does not get sent, and the reason is human. Asking someone for money you have already earned feels like it costs something — the more so if you value the relationship, which means the customers least likely to be chased are often the ones who owe the most. So it waits for a better moment. The better moment does not arrive. Eleven weeks later there is a strained phone call, out of nowhere, about an invoice everyone has forgotten the context of.
That inconsistency is the actual relationship damage — not the chasing. A customer who receives a polite, factual reminder at day seven experiences a business with its house in order. A customer who hears nothing for two months and then gets a tense call experiences something worse, and so does the customer next door who was pursued while they were not. Consistency is what people experience as fair.
So this is close to an ideal automation candidate, with one hard condition. The mechanical part — noticing, timing, drafting, sending, escalating through a ladder you set — should never have required a person. The judgement part — what to do about a dispute, when to stop work, whether to involve anyone else, what to say to someone in genuine trouble — must always be a person, and in this product it is.
The Escalation Ladder — and Its Ceiling
You set the rungs. The AI climbs them and stops. It does not invent a rung above the one you configured.
Before Due — Courtesy
A brief note that the invoice is coming due, with the invoice attached and payment details included. Not a chase — an administrative convenience that catches the wrong-recipient and stuck-in-approval failures before they become lateness.
Just Overdue — Factual Reminder
A polite, factual reminder: invoice number, amount, date issued, days overdue, how to pay. No implication of bad faith, because there almost certainly is none. Timing adapts to that customer’s own payment history rather than a generic rule.
Materially Overdue — Firmer, Still Factual
The tone firms up within limits you set. It states facts about the invoice and asks for a payment date. It does not warn of consequences, threaten action or mention credit reporting — those are claims with legal weight and the AI does not make them.
Handed to a Person
At the threshold you define — by age, amount or relationship — it stops and escalates to a named person with the full history, correspondence and ageing attached. Everything past this point is a commercial decision: stopping work, a payment plan, a collection agency, legal advice.
Any reply from the debtor stops the sequence immediately, at any rung, and routes the conversation to a person. The most damaging thing an automated chaser can do is send reminder four to someone who wrote back three days ago.
What It Actually Does
For your clients’ ledgers — and, if you are honest about your own lockup, for your firm too.
Timing From Real Behaviour
Sequences timed against how that customer actually pays, not a generic template. A customer who has always paid on day 45 is not a risk on day 32 and should not be treated as one.
- Uses the file’s own payment history per customer
- Different sequences by segment, amount and relationship
- Prioritised by value and age
- Fortnightly payment-run cycles accounted for
Tone Under Your Control
Tone is the most configurable thing in the product, because it is the thing that carries the relationship. You set the voice, the firmness curve and the ceiling.
- Your wording, your firmness progression
- Per-client and per-segment tone settings
- Never sarcastic, never implies bad faith
- Invoice and payment details always attached
Any Reply Stops Everything
A reply pauses the sequence for that debtor and routes it to a person, classified to help prioritise — payment claimed, dispute, promise, hardship or query.
- Sequence halts on any inbound reply
- Classified but never resolved by the AI
- Full history and invoice attached on escalation
- Resumes only when a human says so
Conduct Limits Built In
Contact frequency capped and message content bounded. The ACCC and ASIC debt collection guideline and the Australian Consumer Law apply the same whether a person or a system sends it.
- Contact frequency capped by design
- No threats, no legal claims, no credit-reporting warnings
- States facts about the invoice, not consequences
- Full log of what was sent, to whom and when
Ledger Visibility
Ageing, promised dates, disputed items and where each debtor sits on the ladder — visible continuously rather than assembled for a monthly report.
- Live ageing across client files
- Promised payment dates tracked and followed up
- Disputed invoices isolated from the sequence
- Practice-wide view of what needs a person
Run It on Your Own Fees
Firms are notoriously poor at collecting their own fees, for exactly the reasons above. The awkwardness disappears when the reminder is consistent and unemotional.
- Works on your practice ledger as well as clients’
- Partner involved only when there is something to discuss
- Consistent follow-up without the relationship cost
- Feeds your own lockup reporting
The Things It Will Not Say
Every item below is a decision with legal or commercial consequences. Software should not be making any of them.
No threats of legal action
A statement that proceedings will follow is a claim about what your firm or your client intends to do, and it has consequences if it is wrong or if it is never carried out. That is a decision for a person who has considered it, ideally with advice.
No credit-reporting warnings
Warnings about credit listings are consequential, regulated and frequently misused in collections. The AI does not make them. If that step is on your ladder, it is taken by a person who understands the obligations attached to it.
No decisions to stop work or suspend service
Withholding work over an unpaid fee is a commercial and professional decision, with client-relationship and sometimes ethical dimensions. It is escalated to a person, never triggered automatically by an ageing threshold.
No negotiating and no adjudicating
Payment plans, discounts for early settlement, and disputed invoices all stop the sequence and go to a human. A dispute in particular is a signal to stop chasing and start listening — the opposite of what an unattended sequence would do.
This page describes how the product behaves. It is not legal advice on debt recovery — for that, and for anything past the escalation threshold, talk to a qualified adviser.
Related Capabilities
AI Client Query Handling
The inbound counterpart — triage, drafting and the same discipline about escalating anything that needs judgement.
Query handlingAI Bank Reconciliation
Chasing a debtor who has already paid is the fastest way to lose one. Current reconciliation is what prevents it.
Reconciliation prepAI Accounting Cost in Australia
What this kind of automation costs, what drives the range, and how to work out whether it pays for itself.
Cost guideFrequently Asked Questions
Conduct, tone, relationships and where the automation stops.
Sending reminders about money you are genuinely owed is ordinary commercial correspondence, and doing it by software does not change that. What matters is the conduct, not the channel — the ACCC and ASIC publish a joint debt collection guideline setting out what is and is not acceptable, and the Australian Consumer Law prohibits misleading or deceptive conduct, undue harassment and coercion. Those rules apply identically whether a person or a system sends the message. This is precisely why the product is conservative: contact frequency is capped, the messages state facts about the invoice rather than making claims about consequences, and anything that would constitute a threat of legal action, a credit-reporting warning, or a statement about what will happen next is not something the AI sends. Those are decisions with legal consequences, and they belong to a person who has considered them.
Handled badly, it certainly could — which is why tone is the most configurable part of the system and escalation is deliberately conservative. The realistic picture is that most late payment is not defiance, it is administrative: the invoice went to the wrong person, it is stuck awaiting an approval, it was genuinely forgotten. A polite, factual, correctly timed reminder resolves the large majority of those without anyone feeling chased. The relationship damage in practice usually comes from the opposite pattern — inconsistent chasing, where nothing happens for eleven weeks and then a strained phone call arrives out of nowhere, or where one customer is pursued hard and another is not. Consistency is the thing clients actually experience as fair. And the moment a reply carries any friction — a dispute, an apology, hardship, irritation — the sequence stops immediately and a person takes over.
The automated sequence stops for that debtor. Full stop, on any reply. That rule matters more than it sounds: the single worst thing an automated chaser can do is send reminder four to someone who wrote back three days ago explaining that they have paid, or that they are disputing the amount, or that their business is in trouble. Nothing burns a relationship faster than being ignored by a robot. So a reply pauses the sequence and routes the message to a person with the full history and the invoice attached. The AI classifies what it appears to be — payment claimed, dispute, promise to pay, hardship, or a query — to help your team prioritise, but it does not resolve any of them and it does not resume until a human says so.
Yes, and there is a certain justice in it. Accounting firms are notoriously bad at collecting their own fees — partly because the person who should chase is the person who has the relationship, and asking a client for money you have already earned feels like it will cost you something. So it waits, and the firm carries lockup that would be unacceptable in any client it advises. Automating your own follow-up removes the awkwardness in exactly the same way it does for a client’s ledger: the reminder is consistent, factual and unemotional, and the partner only gets involved when there is something genuine to discuss. Firms that run it on themselves first tend to be considerably more confident recommending it to clients.
From your rules and the debtor’s own payment history, not from a generic template. Sequences can differ by client, by customer segment, by amount and by relationship: a long-standing customer who has always paid on day 45 is not a genuine risk at day 32 and does not need a reminder implying otherwise; a new customer drifting past terms on a material amount is a different situation entirely. The AI uses the payment behaviour actually recorded in that file to time follow-up sensibly, and prioritises by amount and age so attention goes where the cash is. What it does not do is escalate on its own initiative beyond the ladder you have configured — the ceiling is set by you, and it does not exceed it.
It does not threaten, it does not make legal claims, and it does not decide anything with consequences. Specifically: no threats of legal action, no warnings about credit reporting, no statements about what will happen if payment is not made, no engagement of a collection agency, no decision to suspend a service or stop work, and no negotiation of a payment plan. Every one of those is a decision with legal or commercial consequences for your firm or your client, and they belong to a person who has weighed them. It also does not adjudicate a disputed invoice — a dispute stops the sequence and goes to a human. What is left is the part that is genuinely mechanical and genuinely valuable: consistent, polite, well-timed follow-up that actually happens.
The Reminder That Actually Gets Sent
Free consultation: bring an aged receivables report — yours or a client’s — and we will map a follow-up ladder you would be comfortable putting your name to. Call +61 3 9999 7398.