Straight answer: AI is brilliant at drafting payment reminders, but the polite, perfectly-worded version it produces by default gets ignored more than a scrappy human one does. Fix the tone (not the tool) and you get paid faster. I tested this on my own invoices over three months and cut my average payment time from 34 days to 19.
The invoice I almost wrote off
Back in early 2025 I had a client, a mid-size events company, sitting on an invoice for £4,200 that was 41 days overdue. I'd sent two reminders. Both had gone through ChatGPT first because I was busy and, frankly, I'd rather spend my energy on client work than wording an awkward email. Both got ignored.
The AI drafts were lovely. "I hope this finds you well." "I wanted to gently follow up." "No rush, but when you get a moment." Every sentence had been softened, rounded off, made palatable. And that's exactly the problem. A busy accounts department that owes forty suppliers money doesn't respond to gentle. They respond to whoever makes the most noise, or whoever sounds like they'll escalate.
So I rewrote the third email myself, kept it short, named the invoice number, named the date, and said plainly that if it wasn't paid within five working days I'd be charging statutory interest and passing it to my accountant to log formally. Paid in three days.
The uncomfortable bit nobody selling you AI templates wants to say
Most "AI for small business" content tells you that AI makes your communication more professional, more consistent, more polished. For invoice chasing specifically, that's the wrong goal. Polished reads as low priority. A big supplier chasing a big client can afford to sound calm because they've got use. A sole trader or a five-person agency doesn't have that use, so sounding calm just gets filed under "not urgent, deal with it next week."
The Federation of Small Businesses has reported for years that late payment is one of the top reasons small firms fail, not lack of sales, lack of cash because the sales they already made haven't been paid for. If your reminder emails are optimised for niceness instead of urgency, you are, without meaning to, training your clients that your invoices sit at the bottom of the pile.
What I changed in my prompts
I didn't stop using AI. I changed what I asked it to do. Instead of "write a polite payment reminder," I now use a three-part prompt structure:
- Tell it the exact facts first: invoice number, amount, date issued, date due, number of days overdue, and any prior contact.
- Tell it the tone explicitly: "direct, brief, no apologising, one clear consequence if unpaid by [date]."
- Tell it what to cut: "no phrases like 'I hope this finds you well,' no hedging words like 'just' or 'gently,' no more than 90 words."
That last instruction matters more than people think. AI models are trained on huge volumes of customer service writing that leans apologetic by default. You have to actively strip that out or you'll get the same soft draft every time, just with different client names slotted in.
The three-email sequence I now use for every client
This is the exact structure, and you can copy it today:
- Day 1 after due date: Short, friendly, assumes it's an oversight. "Invoice #1042 for £X was due on [date]. Could you confirm when this will be settled?" No apology needed, this one is neutral.
- Day 10 after due date: Direct and specific. States the invoice number, the amount, the number of days overdue, and asks for a payment date, not "an update."
- Day 20 after due date: States a consequence plainly: statutory interest under the Late Payment of Commercial Debts Act, work paused on new deliverables, or referral to a collections process. Whichever is true for you, say it, don't imply it.
I built this into a simple spreadsheet with due dates and trigger reminders, and I ask AI to redraft each email using that day's specific numbers rather than writing from scratch each time. It takes about four minutes per client per week.
Where an interactive tool beats another email
One thing that's worked better than any email is showing clients, visually, what late payment is costing them, not me. I built a rough version of this the way Lilach explains in the piece on how interactive calculators pull visitors in and get them to act, except I use it internally: a simple calculator that shows a client what their own supplier is quietly charging them in delayed goodwill, project deprioritisation, and eventually higher rates. Clients respond to a number they can see far more than a paragraph they can skim past.
The data-first approach that changes behaviour
The companies that get paid on time tend to be blunt about numbers rather than vague about feelings. It's the same principle behind how Ahrefs built a brand around being specific and slightly uncomfortable with data instead of hiding behind marketing fluff. When I lead with "this invoice is 41 days overdue and interest is now accruing at 8% plus base rate," there's nothing to interpret. When I lead with "just checking in," there's a whole spectrum of urgency the client gets to invent for themselves, and they'll always invent the lowest one.
Definiteness of purpose beats politeness
Napoleon Hill wrote about definiteness of purpose being worth more than talent, and the same idea from Hill's business lessons applies directly here. A chasing email with a vague purpose ("just wanted to touch base") achieves nothing because it hasn't decided what it wants. A chasing email with one purpose ("pay this by Friday or interest applies") gets acted on because the recipient knows exactly what happens next either way.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
When to stop typing and pick up the phone
AI is useful for drafting, tracking, and keeping your tone consistent across dozens of clients. It is not useful once an invoice hits 30 plus days with no response. At that point you need a phone call, not a fourth email. Howard Schultz talks about this in a different context, having the difficult conversation directly rather than letting it drag through memos, and it's covered well in the piece on Howard Schultz's business lessons. Money conversations are the same. A five-minute call where you say plainly, "I need this paid by Friday, what's stopping that happening?" resolves more than three more polished paragraphs ever will.
Setting this up so it runs without you
If you're running this manually across ten or twenty clients, it's a part-time job by itself. Most small businesses I work with end up automating the tracking (due dates, trigger emails, escalation timing) through their invoicing software or a simple workflow tool, and only bring AI in to draft the actual wording each time so it never sounds copy-pasted. If you want this set up rather than cobbled together in spreadsheets, this is exactly the kind of process an AI consultant for small business can build with you in an afternoon rather than the six months it took me to figure out by trial and error.
The numbers from my own three-month test
Before I changed the tone: average payment time across 22 invoices was 34 days, and four of those went past 60 days.
After switching to the direct, fact-first, no-apology structure: average payment time across 19 invoices was 19 days, and none went past 45 days.
Nothing else changed. Same clients, same invoice amounts, same payment terms on the contracts. The only variable was how the chasing was worded, and whether the consequence was stated rather than implied.
Frequently asked questions
Should I let AI write the whole payment reminder email?
Let it draft, but always strip out apologetic phrases like "just checking in" or "no rush" before sending. Default AI tone is too soft for chasing money, so you need to explicitly instruct it to be brief and direct, or edit that softness out yourself.
How many reminder emails should I send before calling a client?
Two, at most three. Send a neutral one on the due date, a direct one at 10 days overdue, and a final one at 20 days overdue stating a real consequence. Past 30 days with no response, stop emailing and phone them.
Does mentioning statutory interest work, or is it just a threat?
In the UK you're entitled to claim statutory interest on late commercial payments under the Late Payment of Commercial Debts Act, so it's not an empty threat, it's a real right, and stating it plainly tends to move invoices that soft reminders don't.
Is it worth automating invoice chasing if I only have a handful of clients?
If you've got fewer than five recurring clients, a simple spreadsheet with reminder dates is enough. Once you're managing chasing across ten or more clients, the tracking becomes the real time cost, and that's when automating the sequence, with AI only drafting the wording, starts saving you real hours each month.
Related reading: How to Build a Personal Brand as a Consultant in 2026 and The AI Meeting Notes Habit That's Quietly Making Your Team Worse at Their Jobs.
Related: ai payment reminders late paying clients.
