The short version: AI can write your invoice chase emails, tighten your tone, and stop you sending a passive-aggressive message at 11pm when you’re annoyed. It cannot make a client pay you if your payment terms were never enforced in the first place. I tracked this for 90 days across 14 invoices, and the AI-written sequence got me paid on average 9 days faster, but it didn’t touch the two clients who were always going to be a problem, because the problem wasn’t the wording.
Why I Even Started Doing This
Last spring I had £14,200 sitting in unpaid invoices, some of it 60 days old, one of it 94 days old. That last one was a client I liked. Which is exactly why I hadn’t chased it. I’d sent one gentle nudge, got a “so sorry, will sort this week” reply, and then let three weeks slide because chasing someone you like feels rude.
Federation of Small Businesses research has put the average amount owed to a small UK business in late payments at over £20,000 at any given time. That’s not a rounding error. That’s a person not being able to make payroll because someone else is sitting on their money and hoping they’ll forget to ask.
I write about using AI every day without being technical a lot, but I’d never turned it on my own worst habit, which is avoiding uncomfortable money conversations. So I built a chase sequence with ChatGPT and ran it for a quarter.
The Three-Email Sequence I Built
I gave the AI my invoice terms, the client’s payment history, and my actual tone of voice (I fed it three old emails so it didn’t sound like a corporate template). Here’s the sequence that came out, which I still use:
- Day 3 after due date: Friendly, assumes it’s an oversight. “Just flagging invoice #204, due the 12th, in case it slipped past. No drama, let me know if you need anything from me to process it.”
- Day 14: Firmer, names the number and the days. “This is now 14 days overdue at £1,850. Can you confirm a payment date this week?” No apology in this one. AI’s first drafts kept adding “sorry to chase again” and I made a rule: no apologising for asking to be paid.
- Day 30: Direct, mentions consequence. “This invoice is now 30 days overdue. Under the payment terms agreed, statutory interest now applies. I’d rather resolve this by Friday without needing to add that.”
That last line did more work than anything else in the sequence. In the UK you’re legally entitled to charge interest on late business-to-business invoices under the Late Payment of Commercial Debts Regulations, currently 8% plus the Bank of England base rate. Most freelancers never mention it. Most clients don’t know you can. Saying it, calmly, once, changed three conversations out of fourteen.
What Happened After 90 Days
Fourteen invoices went through some or all of the sequence. Here’s the real breakdown:
- 9 paid after the Day 3 or Day 14 email, average 9 days faster than my old habit of one vague nudge and then silence
- 3 paid after the Day 30 email specifically because of the interest line
- 2 didn’t pay on time regardless of what I sent, and both of them are clients I’d already had a gut feeling about before the work even started
That last row is the part nobody selling you an “AI invoice chaser” wants to put in the case study. The two problem invoices weren’t a wording problem. One client had genuine cash flow trouble of their own and no amount of clever phrasing fixes that. The other simply didn’t respect me, and I knew it from the first phone call, when they tried to negotiate my rate down after I’d already sent the proposal. AI can polish the message. It can’t retroactively make you screen the client at the start.
That’s the bit worth sitting with. A better email is worth doing, and worth automating, but it’s a bandage on a wound that opened months earlier when the terms weren’t agreed, or a deposit wasn’t taken, or a scope-of-work document didn’t exist. If you’re consistently chasing the same three clients every single month, the AI email isn’t your fix. Your onboarding process is.
How to Set This Up Yourself This Week
You don’t need software for this, and I’d resist buying a dedicated invoicing-chase app before you’ve tried the free version.
- Open ChatGPT or Claude and paste in your last two or three chase emails, plus one email from a client you find easy to chase and one from a client you find hard. Ask it to identify your natural tone.
- Ask it to write a three-stage sequence (Day 3, Day 14, Day 30) in that tone, with the actual invoice number, amount, and due date left as placeholders you fill in each time.
- Add one line about statutory interest to the Day 30 email even if you never intend to charge it. The mention alone does the work.
- Set calendar reminders for Day 3, 14, and 30 after every invoice due date, or if you use accounting software like Xero or QuickBooks, turn on their automatic overdue reminders and just swap their generic wording for your AI-written version.
- Never send the Day 14 or Day 30 email without re-reading it first. AI drafts can drift too soft or too stiff depending on the day. This is a five-minute edit, not a set-and-forget system.
Total time to build this: about 40 minutes. Time saved per month once it’s running: I’d estimate 2 to 3 hours I used to spend agonising over how to word an awkward chase email, which is exactly the kind of small, boring admin task AI is good at, unlike a lot of the flashier uses people write about.
Where This Falls Down (And Where People Oversell It)
A lot of AI-for-business content treats every use case as a total transformation. It isn’t here. This is a tone-and-timing tool, not a debt collection agency. It won’t chase for you, it won’t call the client, and it definitely won’t make the decision about when to stop working with someone and send it to a solicitor or small claims instead. I made that call myself on the 94-day invoice in the end, after email four went unanswered, and no AI tool should be making that call for you.
I’d also say this, because it’s the part that gets skipped in most “AI for small business” content, including plenty on this very topic: if you’re rebuilding a business after a rough patch, as I have been very publicly over the last few years (I wrote about the month my website started paying me again after a long dry spell), a run of late payments hits differently. It’s not just annoying, it’s frightening, and that fear is exactly why people avoid sending the Day 30 email. The AI doesn’t fix the fear. It just removes the excuse of “I don’t know how to word this,” which for me was about 70% of why I delayed.
If your invoice chasing has turned into a genuine time sink, or you want someone to build the whole cash flow and automation side rather than piecing it together yourself, that’s a fair reason to bring in an AI implementation coach for a few hours rather than muddling through alone. It’s a small, specific job, not a full consultancy retainer, and it should feel that way in the price.
The One Thing I’d Change Doing This Again
I’d send the interest-rate line earlier, at Day 14 rather than Day 30. Waiting a full month to mention it cost me, on reflection, about two extra weeks of chasing on invoices that then got paid within 48 hours of that line landing. It’s the single most useful sentence in the whole sequence and I buried it too late the first time round.
I’d also stop feeding AI content mills the idea that every automation is a silver bullet, the same way I’ve written before about the honest math behind churning out huge volumes of AI content and watching Google index almost none of it. Small, well-aimed use of AI on one specific, recurring, boring task beats a sprawling automation project every time. Chasing invoices is boring. That’s exactly why it’s a good place to start.
Free resource: The Churn Save Email Swipe File.
Frequently asked questions
Can AI get my invoices paid faster?
It can speed up payment from clients who were always going to pay but needed a clearer, firmer nudge, in my case by about 9 days on average. It cannot make a client pay who has no money or no intention to pay, and no chase email, AI-written or otherwise, fixes that.
What should an AI-written invoice chase email include?
The exact invoice number, the amount, the original due date, and by the second or third message, a mention of your legal right to charge statutory interest on late commercial payments. Skip the apology. You’re asking to be paid for work you already delivered.
Is it legal to charge interest on late invoices in the UK?
Yes. Under the Late Payment of Commercial Debts Regulations, UK businesses can charge statutory interest, currently 8% above the Bank of England base rate, on overdue business-to-business invoices, along with a fixed compensation amount. Most small businesses never use this right, which is exactly why mentioning it tends to get a fast response.
What’s a better long-term fix than chasing invoices?
Taking a deposit upfront, usually 30 to 50%, and putting a clear payment schedule and late-payment clause in your contract before work starts. If you’re chasing the same clients every month, the fix sits at the start of the relationship, not in the wording of the reminder email.
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.
If you want your own guide on this published here, you can write for us about finance.
