Straight answer: AI is brilliant at drafting your invoice chasers and tracking who owes you what, but the polished, polite tone it defaults to is the exact reason clients ignore them. What gets you paid is specific numbers, specific dates, and a human name at the bottom, not a well-formatted template. I built a system that uses AI for the boring bits and keeps me in charge of the bit that moves money.
Last year I had 43 invoices go unpaid past their due date. Not one client refused to pay outright. They just didn’t. Emails sat there. Invoices got “lost.” One client told me she’d pay “next week” for eleven weeks running. I run a consulting business, I’ve spent five years rebuilding it after some rough ones, and chasing money is the part of the job nobody warns you about when you go self-employed.
So I built a system with AI doing the heavy lifting on drafting and tracking, and me doing the bit AI can’t do, which is knowing when to stop being nice.
The scale of the problem (it’s not just you)
Federation of Small Businesses research has put the number of UK small firms forced to close each year because of unpaid invoices somewhere in the tens of thousands. Separate industry surveys regularly find the average small business is owed several thousand pounds at any given moment, just sitting in limbo between “invoice sent” and “invoice paid.” If you’ve ever felt mad or embarrassed about having to chase a client for money you’ve already earned, you’re not the problem. The system is built to let people stall.
Here’s what nobody tells you when you go freelance or start a consultancy: the work is often the easy part. Getting paid for it is where the actual friction lives.
What I built (no code, no developer)
I’m not technical and I don’t pretend to be, which I’ve written about in detail on how I use AI every day without any coding background. The invoice chasing system took an afternoon to set up:
- A simple spreadsheet with invoice number, client, amount, due date, and days overdue (a basic formula does this automatically)
- Four email templates drafted with ChatGPT, one for each stage of the chase
- A weekly fifteen-minute slot in my calendar, every Monday morning, to check what’s overdue and send the next stage chaser
- A rule that no invoice goes past 45 days without a phone call, not another email
That’s it. No paid software, no automation platform, nothing complicated. The spreadsheet does the tracking. AI does the drafting. I do the sending and the follow-through.
The four-stage sequence
Stage one goes out the day the invoice becomes overdue. Friendly, brief, assumes it’s an oversight. “Just a quick nudge, invoice 0847 for £1,200 was due Friday, let me know if you need anything from me to get it paid.”
Stage two goes out at day 7. Slightly more direct, restates the amount and the invoice number again (people lose track of which invoice is which when they have several suppliers).
Stage three at day 14 mentions, plainly, that the invoice is now two weeks overdue and asks for a specific date it will be paid by, not a vague “soon.”
Stage four at day 30 is the one that gets results. More on that below, because this is where AI’s default instincts work against you.
The uncomfortable bit: polished AI emails don’t get you paid
Here’s what I learned the hard way. When you ask ChatGPT or Claude to write a “polite but firm invoice chaser,” it gives you something that sounds like every other invoice chaser. Smooth sentences, soft language, phrases like “I hope this finds you well” and “at your earliest convenience.” It reads exactly like what it is: a template. And clients who are stalling on purpose can spot a template from the subject line.
I noticed this when I compared my results. The AI-polished version of my stage-three email got a response rate I’d estimate at maybe one in five. The version I rewrote to be blunter and more specific, invoice number, exact amount, exact number of days overdue, a specific date I needed payment by, got a response almost every time within 48 hours.
This is the same issue I ran into when I wrote about publishing hundreds of AI-drafted blog posts and watching most of them go nowhere. AI content that sounds generic gets ignored, whether that’s a blog post or an invoice chaser, because generic doesn’t demand a response. Specific does.
So my rule now is simple: let AI write the first draft, then strip out every soft phrase and put the hard facts back in. Numbers. Dates. Names. No “at your convenience.” No “just following up.” Say the amount, say the day count, say what you need.
What the day-30 email says
By day 30 I stop asking and start stating. This is the email that got me an £4,200 payment from a client who’d been stalling for eleven weeks:
“Hi Sarah, invoice 0791 for £4,200 is now 32 days overdue. I’ve resent the invoice and my bank details are attached again in case the original went astray. I need this settled by Friday 14th to keep my own accounts straight this month. If there’s a genuine issue with the invoice, please tell me today so we can sort it, otherwise I’ll expect payment by Friday.”
No apology for chasing. No softening. A specific deadline. An offer to fix a real problem if there is one, which closes off the “I never got it” excuse politely but firmly. She paid on the Thursday.
Where AI helps and where it doesn’t
AI is excellent at three things in this whole process: drafting the first version of a chaser fast, checking tone before you hit send (I regularly paste an email in and ask “does this sound too apologetic or too aggressive”), and summarising a client’s payment history so you can see patterns, like the client who always pays late in December, or the one who only responds after the third email.
AI is bad at one crucial thing: knowing your relationship with that client. It doesn’t know that this particular client is going through a rough patch and deserves a bit of grace, or that this other one has stalled you three times running and needs the phone call, not another email. That judgement call has to stay with you. I use AI to write the words. I decide when to send them and when to pick up the phone instead.
The spreadsheet trick that changed everything
The single biggest improvement wasn’t a tool, it was a habit. I stopped checking for overdue invoices “whenever I remembered” and put it on the calendar as a fixed Monday morning task, same as I would check analytics after my website started earning again after a long SEO recovery. Money owed to you needs the same discipline as money you’re trying to earn. Fifteen minutes every Monday, spreadsheet open, chasers sent, no exceptions.
Within four months of running this, my average days-to-payment dropped from around 38 days to 19. That’s not a huge sample size, but it’s my real numbers across 43 invoices, and the pattern held steady enough that I trust it.
A few practical rules I stick to now
- Never let an invoice go past 45 days without a phone call. Emails get “lost.” Phone calls don’t.
- Always restate the invoice number and exact amount in every single chaser, even the first one. Clients juggle several suppliers and forget which invoice is which.
- Cut every soft phrase AI gives you by default. “I hope this finds you well,” “at your earliest convenience,” “just a gentle reminder” all get deleted.
- Give a specific date, never “soon.” “By Friday the 14th” gets a response. “At your earliest convenience” gets ignored for another fortnight.
- For repeat late payers, ask for a deposit or 50 percent upfront on the next job. Their payment history is the best predictor you have.
There are also plenty of free tools worth knowing about if you want to go beyond a spreadsheet, some of which I’ve come across while researching odd corners of the internet for a completely different piece on websites you probably didn’t know existed, several of which turned out to be useful invoice and payment tracking tools for tiny businesses that don’t want to pay for full accounting software.
What I’d tell anyone starting this from scratch
Don’t outsource the judgement, only the drafting. Set a fixed weekly slot and treat it as non-negotiable, the same way you’d treat a client meeting. And accept the uncomfortable truth that being nicer doesn’t get you paid faster. Being specific does. If you want a system that’s built around genuine numbers and human judgement rather than automation for its own sake, that’s the whole approach worth copying, whether it’s chasing invoices or anything else in the business you’re trying to run more efficiently. If you’re weighing up whether it’s worth getting outside help to build systems like this across your whole business, it’s worth reading about what an AI consultant costs before you decide whether to DIY it or bring someone in.
Frequently asked questions
Can AI write my invoice chasing emails for me?
Yes, and it’s a good starting point for structure and tone, but the first draft AI gives you will almost always be too soft and too generic to get a response. Use it to write the skeleton, then add back the exact invoice number, amount, days overdue, and a specific date, and cut anything that sounds like a template.
How many days should I wait before chasing an unpaid invoice?
Send the first reminder the day it becomes overdue, not weeks later. Follow up again at day 7, day 14, and day 30. By day 45, stop emailing and call instead. Waiting longer than that to chase makes clients assume you’re not tracking it closely, which makes them deprioritise you further.
Is it rude to chase a client for payment?
No. You’ve delivered the work, the invoice is a contract, and chasing payment on time is a normal part of running a business, not a personal favour you’re asking for. The politeness that stops people chasing is exactly what lets clients stall for months.
What’s the single biggest mistake small businesses make with late payments?
Treating it as an occasional annoyance instead of a fixed weekly task. Without a scheduled check, overdue invoices quietly slide from 10 days to 40 days because nobody is looking at them on a set day. A fixed fifteen-minute slot every week fixes most of the problem before it becomes a crisis.
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.
