The short version: I used AI to build a three-stage invoice-chasing sequence for eight overdue clients over 60 days, and six of them paid faster than usual. But the AI didn’t fix late payment, my own habit of sending vague terms and chasing too politely did. The tool only worked once I fixed the process underneath it.
Worth reading next: AI-Written Proposals Are Losing You Clients (Here’s What to Send Inste.
The invoice that started this experiment
Forty-seven days. That’s how long a marketing agency I worked with (I’ll call them Client A, because naming them would be petty and I’ve got better things to do with my Tuesday) sat on a £4,200 invoice before I did anything about it beyond one gentle email and a lot of hoping.
I run a business. I know exactly how this looks from the outside. I coach clients on cash flow and payment terms and here I was, letting my own invoice go unchased because I hate that particular conversation. It feels awkward asking for money you’re owed, even when the work is done, the client’s happy, and the contract is clear. Most small business owners I talk to feel the same way. It’s not a skills gap, it’s a discomfort gap.
So I decided to test whether AI could take the discomfort out of it for me.
What I built (and how simple it was)
I fed ChatGP a copy of my standard invoice terms, the payment history for eight overdue accounts, and asked for three versions of a chase sequence: friendly, firm, and formal. Nothing clever. Here’s the sequence I ended up running for 60 days:
- Day 1 after due date: A short, warm reminder assuming it’s an oversight. No mention of consequences.
- Day 14: A firmer note referencing the specific invoice number, amount, and days overdue, plus a mention of statutory late payment interest under the Late Payment of Commercial Debts Act, which lets UK businesses charge 8% plus the Bank of England base rate on overdue B2B invoices.
- Day 30: A formal letter tone, referencing the earlier two emails by date, restating the amount owed, and offering a payment plan as the last soft option before anything more formal.
Before this system, my average time to payment across those eight clients was 41 days. After running the sequence, it dropped to 19 days for six of them. Two didn’t budge at all until I picked up the phone, which tells you something important I’ll come back to.
The uncomfortable part nobody tells you
Here’s what surprised me. I sent the Day 14 email, word for word as the AI wrote it, to a client I’d worked with for three years. She called me the next day, not angry exactly, but hurt. She said it read “like a template” and asked if I’d sent it to other people too. Which I had. She wasn’t wrong to notice.
That’s the bit most people writing about AI and business admin skip over. AI-written chasing emails work brilliantly on new clients, one-off projects, and anyone you don’t have a relationship with. On people you’ve worked with for years, the same polished, slightly generic tone reads as cold, and it can do more damage to the relationship than the late payment ever did. I had to ring her, apologise, and explain I’d been testing a new system, which was true and still a bit embarrassing.
The lesson wasn’t “don’t use AI.” It was: use AI to draft, then edit in the specifics that prove a human wrote it. A reference to a call you had, a project detail, her dog’s name if that’s your relationship. Thirty seconds of personalising turned a template into a message that still felt like me.
The real fix was never the chasing email
Here’s the truth I didn’t want to admit at the start of this. AI made me faster at chasing, but it didn’t fix why I was chasing in the first place. My invoices didn’t always state payment terms clearly at the point of sending. Some went out with no due date at all, just “payment appreciated,” which is basically an invitation to pay whenever suits. Once I started putting a hard due date, the interest clause, and a link to my payment page on every single invoice before the work even finished, three clients I’d expected to chase paid on time without a single reminder.
AI speeds up a broken process. It doesn’t repair one. That’s a distinction worth sitting with if you’re about to automate your way out of a habit you haven’t addressed.
The behavioural trick that did more than the wording
The bit that made the biggest difference wasn’t tone, it was timing and framing, which is straight out of behavioural economics. I’ve written before about the business lessons from Richard Thaler, and his work on nudges applies directly here. People respond to defaults and to loss framing more than they respond to politeness. Telling a client “you’ll be charged 8% plus base rate interest from day 14” is a nudge, not a threat, and it consistently got faster replies than the friendliest wording I could write.
If you want to take this further, build a simple calculator on your invoice page showing a client exactly what their bill grows to for every week it’s late. It’s the same principle behind the interactive calculators piece I wrote a while back, numbers that update in real time make consequences feel real in a way a paragraph of text never quite does.
Tone still matters more than the technology
None of this works if the tone is aggressive. I looked at how bigger brands handle sensitive messaging for inspiration, and Headspace is a good example, their whole brand is built on staying calm even when the subject (anxiety, stress, sleep) is uncomfortable. I wrote about their approach in the Headspace marketing strategy breakdown, and the same discipline applies to chasing money. Calm and firm beats sharp and defensive every time, even at Day 30.
Small business owners aren’t debt collectors, and clients know it
One thing worth remembering: you’re not a collections agency, and sounding like one damages trust fast. GoDaddy built an entire brand around speaking to small business owners like humans rather than corporations, which I covered in the GoDaddy marketing strategy piece. That same plain, human register is what kept my Day 1 and Day 14 emails from sounding like threats. The formal tone only came out at Day 30, and only once.
What I’d tell you to do this week
If you’re sitting on unpaid invoices right now, here’s the version of this I’d run if I were starting again:
- Write your three-stage sequence with AI as a first draft, then rewrite the middle paragraph of each one by hand so it doesn’t read as a template.
- Put a hard due date and the statutory interest clause on every invoice you send from today, before you send another chase email.
- Set calendar reminders for Day 1, Day 14, and Day 30 after the due date, don’t rely on remembering.
- Pick up the phone for anyone who’s been a client for more than a year. The email opens the conversation, the call closes it.
- Track how many days it takes each client to pay, so you know who needs shorter terms next time.
If admin like this is eating hours you don’t have, it’s worth setting up once with help rather than rebuilding it every quarter. This is exactly the kind of repeatable system an AI implementation coach can help you set up so it runs itself instead of living in your head as a dreaded Tuesday task.
Frequently asked questions
Can AI legally chase invoices for me in the UK?
Yes, there’s no legal restriction on using AI to draft chasing emails, the content just needs to be accurate. Under the Late Payment of Commercial Debts Act 2013, UK businesses can charge statutory interest of 8% plus the Bank of England base rate on overdue B2B invoices, and it’s worth mentioning this in your later-stage emails, whether AI drafted them or not.
Does AI-written chasing work better than doing it yourself?
It works better at getting you to send the reminder at all, since most people delay chasing out of discomfort rather than lack of skill. The wording itself performs similarly to what most people would write anyway. The main gain is consistency, not persuasion.
Will clients notice if my invoice emails are written by AI?
Long-term clients often will, especially if the tone shifts noticeably from how you normally write to them. The fix is to use AI for structure and drafting, then personalise the middle of each email with a specific detail only a human who knows the client would include.
What’s the fastest way to reduce late payments without chasing at all?
Put a firm due date, payment link, and interest clause on the invoice itself at the point you send it, rather than adding these after the due date has passed. In my own test, invoices sent with clear terms upfront got paid on time far more often than ones I had to chase afterwards.
Want this done for you? See whether to hire an AI consultant or an agency.
Related reading: Away Marketing Strategy: How They Built a Brand That Wins and Dollar Shave Club Marketing Strategy: How They Built a Brand That Wins.
