Debtor Delay Tactics in 2026: The AI Stalling Playbook (and How to Beat It)
Somewhere right now, a clerk in your debtor's finance department is typing four words into a chatbot:
"Write a polite reply."
Ten seconds later, you receive a beautifully written letter explaining why your invoice can't be paid this month.
Debtor delay tactics are as old as invoices. What changed in 2026 is the cost of running them. Stalling used to take effort. Now it takes a prompt.
This guide covers the five delay tactics that AI has supercharged, how to recognise each one, and the exact responses that cut through them.
Why debtor stalling got cheaper in 2026
We're not guessing about AI adoption in finance departments. Anthropic โ the company behind the Claude AI models โ publishes the Economic Index, a dataset mapping real AI usage to workplace tasks across 121 countries.
Three findings matter for anyone chasing unpaid invoices:
- Admin and finance tasks are heavily AI-touched. Office & administrative support tasks make up about 7.9% of all observed AI usage worldwide. Business & financial operations tasks add another 5.8%.
- Emails are a top output. "Email or message" runs above the global average as an AI output in the US, UK, France and Germany. The letter is the product.
- Nearly half of AI use is full delegation. Globally, about 49% of observed usage is "automation" โ the task handed over entirely, not just assisted.
One honest caveat: this data measures tasks matched in AI conversations, not people. It can't tell you what percentage of AP clerks use AI. But it tells you the capability is there, on the desk, today.
The Back Office Went Digital
AI usage matched to workplace tasks โ worldwide, May 2026
Source: Anthropic Economic Index, May 2026 snapshot
The 5 debtor delay tactics AI made stronger
1. The perfect dispute letter
The tactic: A sudden quality complaint about goods delivered months ago, written in flawless commercial English, quoting your own contract terms back at you.
What changed: A convincing dispute used to signal the debtor had escalated internally, maybe involved a lawyer. Now it signals nothing. Anyone can generate legally literate prose in seconds.
Your move: Stop reading the writing quality as a risk signal. Ask one question instead: why was this defect never mentioned before the invoice fell due? Demand contemporaneous evidence โ inspection reports, rejection notices, emails from the delivery week. AI can write a complaint. It cannot backdate one.
2. The infinite reply loop
The tactic: Every letter you send gets a prompt, courteous, substantive-looking reply. Nothing is refused. Nothing is paid. The correspondence simply never ends.
What changed: Replying used to cost the debtor time โ and eventually they'd rather pay than keep writing. That pressure is gone. Reply capacity is now unlimited and free.
Your move: Escalate on the calendar, not on the tone of the replies. Set your deadline at day one, and when it passes, move โ regardless of how reasonable the last letter sounded. A debtor who wants to pay sends money, not paragraphs.
3. The moving-target contact
The tactic: "This needs sign-off from our finance director, who is travelling." Next month it's the auditors. Then year-end close. The person who can actually pay is permanently unavailable.
What changed: AI drafts each handoff smoothly and consistently, so the story never contradicts itself โ the classic way this tactic used to unravel.
Your move: Name the human. Ask, in writing, for the name and title of the person authorised to approve payment, and address everything to them. A language model can write on behalf of a department. It struggles when a named officer must personally own the refusal.
4. The precision part-payment
The tactic: A small payment arrives โ often an oddly specific amount โ with a note that the "remainder is under review." It resets your patience and, the debtor hopes, your escalation clock.
What changed: AI makes it trivial to calculate exactly how little keeps a creditor calm, and to wrap the amount in a credible-sounding review narrative.
Your move: Accept the money, reject the story. Confirm receipt in writing as partial payment of the undisputed full amount, restate the balance and the original deadline, and keep the escalation date unchanged. Part-payment is evidence the debt is acknowledged โ use it as leverage, not a reason to wait.
5. The jurisdiction fog
The tactic: Hints that the file is "with our legal team" to review governing law, or vague references to local formalities that supposedly delay everything.
What changed: Debtors can now research your enforcement options as fast as you can โ and craft delay narratives around the genuinely slow parts of your legal route.
Your move: Get ahead of it. Know your realistic enforcement timeline in the debtor's country before the first reminder, and reference it specifically. Fog only works on creditors who haven't checked the map.
The Excuse Decoder
What the letter says โ and the reply that works
What still works when the letters are machine-written
Here's the good news: AI changed the debtor's correspondence, not their obligations. And three creditor moves have actually become more effective, because they target exactly what automation can't fake.
Specificity. Reference the named signatory on the delivery note, the exact shipment, the precise consequence arriving on a precise date. Generic pressure gets absorbed by generic replies. Specific pressure demands a human answer.
The telephone. A chatbot writes the letter. It doesn't take the call. Voice contact remains the fastest way to find out whether you're dealing with a dispute or a delay.
The calendar. The single biggest AI-era mistake creditors make: letting polished replies reset their internal clock. Fix your escalation date on day one. Nothing in the correspondence moves it โ only money does.
The Fixed-Clock Escalation
Set once. Replies don't move it. Only payment does.
Frequently asked questions
How can I tell if a debtor's letter was written by AI?
Increasingly, you can't โ and it doesn't matter. Treat every reply the same way: check whether it contains a payment date and evidence, and ignore the polish. Behaviour is the signal now, not prose.
Should I use AI to write my own collection letters?
For structure and speed, yes. But keep the specifics human: real names, real shipment references, real dates. A generic AI letter to a debtor with a generic AI inbox is two machines wasting each other's time.
Do delay tactics differ by country?
Yes, and AI adoption differs too. The Anthropic Economic Index ranks France (#10) and the US (#12) well ahead of Germany (#26) in overall AI usage relative to population โ so expect more automated correspondence from some markets than others, and calibrate your timelines to each jurisdiction's real enforcement route.
When should I hand an unpaid invoice to a collection agency?
When the fixed clock says so โ typically day 45โ60 for B2B invoices โ or immediately if the debtor stops responding, disputes without evidence, or shows insolvency signals. Every week of extra correspondence lowers recovery odds.
Stop answering the machine
Your debtor automated their excuses. The professional answer isn't better letters โ it's taking the file out of the correspondence game entirely.
Intercol collects B2B debts in the debtor's own jurisdiction, in the debtor's own language, on a no-collection, no-fee basis.
