The short version: AI can write you a chasing email that is firm, polite, and doesn't sound like you've been up at 2am fuming about it, but it will not make someone pay you who never intended to. I ran a three-email AI sequence on my own overdue invoices for three months, got paid an average of nine days faster, and lost exactly one client, who was going to be a problem client anyway.
Why I even started doing this
Two years ago, rebuilding my consultancy after a rough patch, I had a spreadsheet that made me feel slightly sick every time I opened it. Four invoices, all over 30 days late, all from clients I liked, which is exactly why I hadn't chased them. Liking someone is the worst reason to let them owe you money for six weeks.
The Federation of Small Businesses has been saying for years that the average UK small business is sitting on five figures of unpaid invoices at any given moment. I believe it, because I was one of those businesses. Not because clients were malicious. Because I was too polite, too slow, and honestly a bit embarrassed to ask twice.
So I did what I do with most operational problems now: I handed the awkward, repetitive part to AI and kept the judgment calls for myself.
The client this happened with
I'll call her Chloe, because that's not her name. Chloe ran a small events agency, three staff, lovely woman, terrible payment habits. Not dishonest, just chaotic. She'd paid me late twice already for a marketing project, always with an apology and always eventually, but "eventually" was costing me cash flow I didn't have spare.
Invoice three went out. Thirty days passed. Nothing. I asked ChatGPT to draft a reminder using her name, the invoice number, the original due date, and a tone brief: "professional, warm, no anger, assume good faith, but make the next step clear." It gave me something better than what I'd have written at 11pm annoyed, because annoyed-me writes passive-aggressive sentences that sound like a customer service complaint, not a business owner.
She paid within four days of that email. The old me would have sent a vague "just checking in!" message and waited another two weeks for a reply.
The three-stage AI sequence I now use for every late invoice
This isn't complicated. It's three emails, spaced out, each with a slightly different job. I keep the prompts saved so I'm not starting from scratch every time.
- Email one, day 1 to 3 late: Friendly nudge. Prompt: "Write a short, warm reminder that invoice [number] for [amount], due [date], hasn't come through yet. Assume it's an admin oversight. Include the payment link and ask them to flag if there's a query." This one almost never fails, and it's the one most small business owners skip entirely because they don't want to seem pushy at three days.
- Email two, day 14 to 21 late: Firmer, no apology in the tone, no exclamation marks. Prompt: "Write a follow-up reminder for the same invoice, now [X] days overdue. Reference the previous email, state the amount owed clearly, and ask for a payment date, not just a promise to look into it." This is where I stop using soft words like "just" and "quick" that I'd have used naturally.
- Email three, day 30 plus: This is where AI drafts, but I personally edit hard. Prompt: "Write a final reminder before this goes to a formal late payment process, referencing the Late Payment of Commercial Debts Act, still professional, no threats, but clear about the next step and the date." I always read this one twice before sending, because it's the email that ends relationships if it's clumsy.
Three emails, roughly 400 words total, took me under ten minutes across three months to draft and send for four different clients. The old way took me longer because I'd draft it, delete it, feel guilty, and rewrite it softer, which is exactly the wrong direction.
What moved faster and by how much
I tracked payment dates before and after I started using this sequence. Across eleven late invoices over three months, average time from due date to payment dropped from 24 days to 15 days. Nine days doesn't sound dramatic until you're a one-person consultancy and that's nine fewer days your own bills are waiting on someone else's cash flow decision.
One invoice didn't get paid at all, from a client who'd clearly decided not to pay before I sent anything. No email sequence, AI-written or otherwise, fixes a client who never intended to settle up. That's the uncomfortable bit nobody selling you an invoicing tool wants to say out loud: some percentage of your late payers were never going to pay, and being politer to them for longer just delays the moment you write it off or go legal. AI made me faster at spotting which category someone was in, because a templated, well-worded chase either gets a payment date within days or it gets silence, and silence after three professional emails tells you everything.
The part AI can't do for you
It can't decide your payment terms. It can't decide whether you take a deposit upfront. It can't make you the kind of business owner who states, in the contract, exactly what happens at day 30, day 45, day 60. I learned the hard way that the businesses I struggled to get paid by were almost always the ones where I'd been vague about terms at the start, because I didn't want to seem difficult during a sales conversation.
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.
Now every proposal I send states payment terms in plain numbers, not "prompt payment appreciated." AI writes my chasing emails well because the groundwork, the actual terms, is already clear. Feed it a fuzzy contract and it'll write you a fuzzy, ineffective reminder too, because there's nothing firm underneath the polite words.
How to set this up in your own business this week
- Pull every invoice currently over 14 days late into one list, with invoice number, amount, and days overdue.
- Write one prompt template for each of the three stages above and save them somewhere you'll reuse them, a notes doc, a saved prompt in ChatGPT, whatever sticks.
- Set a recurring 15-minute slot, weekly, to check what's newly overdue and run it through the sequence. Chasing invoices in a scheduled slot removes the emotional decision of "should I chase today," which is the thing that quietly costs small businesses thousands.
- If you use invoicing software like FreeAgent, Xero, or QuickBooks, most already send automated day-one reminders. Layer the AI-drafted personal follow-up on top from day 14 onward, because that's the point where a template email from your accounting software starts getting ignored.
- Decide your day-60 line in advance, before you're emotionally in it. Mine is a formal letter before debt collection or small claims. Deciding that line calmly, in advance, means email three doesn't waver.
If setting up this kind of workflow across your invoicing, follow-ups, and general admin sounds like exactly the sort of thing you'd rather have someone build once and hand to you, that's a fairly typical starting project when people bring me in as an AI implementation coach, alongside content and lead generation work.
Where the bigger money problem sits
I'll be honest, chasing invoices faster is a fix at the shallow end. The deeper issue for most small businesses I talk to is that they're not generating enough new leads to make one slow-paying client feel manageable rather than catastrophic. If your pipeline is thin, a single late invoice becomes a five-alarm fire. Fixing the marketing side, the actual flow of new enquiries, matters more long term than any email template, which is part of why I've written before about tools like interactive calculators for converting website visitors, because more qualified leads coming in steadily is the real cushion against one client sitting on your money for six weeks.
It also connects to something I learned rebuilding my own site traffic after a rough patch, documented in the month my website started paying me again: cash flow problems in a small business are rarely one single issue, they're usually three smaller issues stacked, thin pipeline, slow payers, and no system for either. AI fixes the second one nicely. It won't fix the first for you.
A quick note on tone, because it trips people up
The biggest mistake I see when people first try AI-written chasing emails is asking for "assertive" and getting something that reads like a warning letter for invoice one. Save the assertive tone for email three. Email one should read like you think it's an oversight, because most of the time, at that early stage, it is. If you skip straight to firm, you'll damage a client relationship over something that was probably just a Friday afternoon admin backlog on their end. I learned this by overcorrecting after Chloe, sending a slightly too-sharp first email to a completely different client who'd simply had a member of staff off sick. She told me later it stung. Fair enough. I softened stage one permanently after that.
Frequently asked questions
Will AI-written invoice reminders sound robotic or generic to clients?
Not if you feed it specifics: the client's name, the exact invoice number, the actual due date, and a clear tone instruction. Generic prompts produce generic emails. Specific prompts, with real details from your invoice, produce something that reads like you wrote it, because in the parts that matter, you did.
How many days late should the first AI reminder go out?
Send it the day after the due date, or within a day or two at most. Waiting a week to send the first nudge because you don't want to seem pushy is the single most common mistake small business owners make, and it's the easiest one to fix once you've got a template ready to go.
What if the client still doesn't pay after three AI-drafted emails?
That's your signal to stop being polite and start being formal. In the UK, the Late Payment of Commercial Debts (Interest) Act gives you the right to charge statutory interest and a fixed compensation fee on business-to-business late payments. At that point you're either sending a formal letter before action or moving to small claims, and no amount of well-worded reminders will substitute for that decision.
Is it worth paying for dedicated invoice-chasing software instead of just using AI to draft emails?
If you've got more than a handful of clients and can't stay on top of a weekly review, tools like Chaser can automate the whole sequence for you. For most small, one-to-five person businesses, a saved set of AI prompts and a 15-minute weekly habit does the same job for free, which is where I'd start before paying for anything.
Related reading: The AI Subscription Stack: What I Pay For (and What I Cancelled) in 2026 and I Got AI to Chase My Late Invoices. Here's What Happened.
