How to automate invoice reminders

You sent the invoice three weeks ago. The work was good. The client was happy. And yet the money isn't in your account, and now you've got a decision to make that you'll put off until Thursday: do you send the awkward email, or do you wait another week and hope?

Multiply that by the four or five invoices that are always quietly overdue, and you've got the single most common admin drain on a small business. The good news is that almost all of it is a rules problem, not a judgement problem — which means it's the kind of work you can hand off to a machine and stop thinking about. Let's walk through exactly how, and where the one small dose of AI actually earns its keep.

The real cost of chasing invoices by hand

Late payments are a cash-flow problem dressed up as an admin problem. On paper you're owed the money; in reality it's not in the bank, and you can't pay a supplier or yourself with money you're owed. A business with £18,000 sitting in overdue invoices isn't a rich business having a slow week — it's a business one bad month away from a genuine squeeze, quietly funding its clients' cash flow for free.

Then there's the relationship damage, which is the part people underestimate. Chasing feels rude. So you soften it, delay it, or word it so gently the client doesn't register it as a chase at all. Or you finally snap after a month of silence and fire off something sharper than you meant, and now a good client feels told off over an invoice they'd simply lost in their inbox. The awkwardness isn't the payment — it's that you have to be the one to raise it, again, as a person they know.

And underneath both of those sits the mental load. Keeping a running tally in your head of who owes what, how overdue it is, whether you already chased them, and whether that gentle nudge two weeks ago ever got a reply. It's a background process that never fully shuts off. You're in the shower and you remember the Henderson invoice. That's not a to-do; it's a low hum of anxiety that follows you around. Getting rid of that hum is a big part of what actually gives a small team its Friday back — not the twenty minutes of typing, but the hours of remembering.

Why this is an automation job, not an AI one

Before building anything, it's worth being honest about what kind of tool the job needs. We use a rough five-minute test, and invoice chasing passes it with room to spare. The test asks three questions.

Is the input tidy? Yes. Every invoice already has an amount, an issue date, a due date, a client, and a paid/unpaid status. That's structured data sitting in your accounting tool. No messy free-text to interpret, no photos to read, no tone to gauge.

Can you write the rule down? Yes, on the back of a napkin. "If an invoice is unpaid and the due date has passed, send reminder one. Seven days later, if still unpaid, send reminder two." A rule you can state in a plain sentence is a rule a machine can follow exactly.

Does it have to be exact? Absolutely. You cannot have the system chase a client who paid yesterday, or send a "firm" reminder when it should have sent a gentle one, or get the amount wrong by a penny. This is the clincher. AI is brilliant at fuzzy, judgement-heavy tasks where "close enough" is a win — and dangerously wrong for tasks where the answer must be precise every single time. An invoice chase must be precise every single time.

So the verdict writes itself: this is a plain automation. If you want the longer version of how we decide, we wrote it up in AI vs automation — how to tell which one you actually need. The short version: reach for the boring, deterministic tool for anything that runs on tidy data and firm rules, and save the clever, generative stuff for jobs that genuinely need to think.

The reminder ladder

The heart of the system is an escalation schedule — a ladder. Each rung is triggered purely by how many days an invoice has been unpaid, and the tone rises as you climb. Here's a schedule that works for most small businesses on standard 14- or 30-day terms.

• Due date (day 0): a neutral, friendly heads-up. "Just a note that invoice #1043 is due today." No accusation, because nothing's wrong yet. Roughly a third of clients pay on this nudge alone, simply because it lands the payment back on their radar.
• Day 7 overdue: the first genuine reminder. Still warm, but clear that the date has passed. "This one's now a week overdue — could you let me know when it's likely to be settled?"
• Day 14 overdue: firmer, and this is where you restate the specifics. Amount, invoice number, original due date, and a direct ask for payment or a date. The politeness stays; the vagueness goes.
• Day 30 overdue → human: the machine stops here and hands the invoice to you. Thirty days of silence means something is actually wrong — a dispute, a cash-flow problem at their end, a lost invoice, or a client going quiet. That needs a phone call from a person, not a fourth automated email. The automation's job is to make sure this invoice lands on your desk flagged and impossible to miss, not to keep nagging into the void.

You can tune the intervals — some businesses go 3/10/21, some slower — but the shape holds: quiet nudge, real reminder, firm reminder, then a human. Four rungs, then a person.

Writing reminders that get paid without burning the relationship

The wording matters more than the schedule. A well-worded reminder gets paid and keeps the client; a badly worded one gets paid once and quietly loses you the next job. You need three templates, mapped to the ladder.

The gentle one (due date / day 7). Assume the best. People aren't dodging you; they're busy and it slipped. Keep it short, keep it warm, make paying easy — the invoice attached, the amount, and the payment link right there. "Hi Sarah, hope the new office is settling in. Quick note that invoice #1043 (£1,200) is due today — link below whenever it's convenient. Thanks!"

The friendly-but-clear one (day 14). Still human, but now you state the facts plainly and ask a direct question. "Hi Sarah, following up on invoice #1043 for £1,200, which was due on the 5th and is now two weeks overdue. Could you let me know when I can expect it, or if there's anything holding it up?" That last clause matters — it opens a door for them to tell you about a problem rather than just go quieter.

The firm one (day 21–30). Polite, unambiguous, and it names the next step. "Hi Sarah, invoice #1043 (£1,200) is now three weeks overdue. I'd appreciate payment by Friday. If there's an issue with the invoice or the timing, please do call me so we can sort it out." Firm isn't rude. Firm is clear about what you need and by when.

A good reminder assumes the best about the reader and is completely clear about the ask. Warmth and precision aren't opposites — the awkward chases are the ones that are vague about both.

Three tone principles run through all of them: never accuse (the invoice was probably just missed), always make paying frictionless (link and amount in every message), and always keep the door open for a conversation. You're chasing money, not winning an argument.

How to set it up on what you already have

Here's the part that surprises people: you almost certainly don't need to buy anything. The whole system runs on data you already keep.

If you use accounting software (Xero, QuickBooks, FreeAgent, Wave and most others), it already tracks issue date, due date, amount and paid status — and several have automated reminders built in, sitting switched off in a settings menu you've never opened. Step one is genuinely just: find that setting, write your three templates in, set the intervals, turn it on. Ten minutes. For a lot of businesses that's the entire project, and there's nothing to build.

If you invoice from a spreadsheet, you're still fine. A sheet with columns for client, amount, due date, paid (yes/no) and last-chased date has everything the rules need. A light automation layer — a scheduled script or a no-code tool — reads the sheet each morning, works out which invoices are on which rung, and sends the matching template. The spreadsheet stays your source of truth; the automation is a thin worker sitting on top of it.

If your invoicing tool can't do it and you want it properly hands-off, that thin automation layer is a small, cheap build — connect to wherever the invoice data lives, apply the ladder, send the email, log what it did. This is deliberately boring software, and boring is exactly what you want for something that touches money. If you're weighing up a build, it's the sort of tightly-scoped job we'll quote a fixed price for, precisely because the rules are so well-defined there are no surprises to hide behind.

The one place a pinch of AI can help

Being pro-right-tool doesn't mean being anti-AI — it means putting it where it actually fits. And there's exactly one spot in this whole system where a language model genuinely earns its place: writing the templates, once, up front.

Staring at a blank page trying to word a firm-but-friendly chase is real work, and it's the fuzzy, judgement-flavoured task AI is good at. Ask it to draft three reminders — gentle, friendly, firm — in your voice, for your kind of client, and it'll give you a solid first draft in seconds. You then edit them to sound like you, and you're done.

Notice what the AI is not doing: it isn't deciding who to chase, when to chase, or whether an invoice is paid. It drafts the words once; the deterministic automation runs the chase forever. Let AI help you write the templates, and never let it near the send button. That division — clever tool for the one-off creative bit, boring tool for the repeated exact bit — is the whole philosophy in miniature.

Handling the weird 20%

Most invoices are simple: unpaid, then paid. But a fraction go sideways, and a good system knows the difference between the 80% it should handle and the 20% it must not touch.

• Disputes. "I'm not paying because the work wasn't right." An automated firm reminder here pours petrol on a fire. The moment a client replies contesting an invoice, the automation must stop chasing that invoice and flag it to you.
• Partial payments. They've paid £600 of £1,200. Is that paid? Half-paid? Marking it "paid" stops the chase on money you're still owed; marking it "unpaid" chases someone who's genuinely paying. This needs a human to look.
• Payment plans. You've agreed they'll pay in three instalments. The generic due-date rule doesn't know that and will chase them for being "overdue" on a schedule you personally agreed. This invoice needs pulling out of the standard ladder entirely.

The design principle is simple: automate the 80% that's clean, and route the exceptions to a person. The goal isn't a system that never needs a human. It's a system that only needs a human when it should — freeing you from the fifteen easy chases so you've got the attention for the three hard ones that actually need you.

A worked example, with the maths

Take a small studio sending about 40 invoices a month. Roughly a third pay late enough to need chasing — call it 13 invoices. Each one, done by hand, is maybe 10 minutes: find the invoice, check it's still unpaid, decide the tone, word the email, send it, note that you did. Often it's more, because you also spend time deciding whether to chase at all.

13 invoices × 10 minutes ≈ 2 hours 10 minutes a month of pure chasing admin. Call it 26 hours a year. If your time is worth £50 an hour, that's around £1,300 a year in your own time spent typing awkward emails — before you count the ones you forgot to send.

Those forgotten ones are where the real money hides. When chasing is manual, plenty of invoices simply don't get chased on time, and every extra week an invoice sits unpaid is a week you're financing a client for free. If automating the chase pulls your average payment time down from, say, 45 days to 30 — a completely realistic swing once every invoice gets a reliable day-0 nudge — a business turning over £200,000 a year frees up something in the order of £8,000 of cash that was previously stuck in the "owed but not received" column at any given moment.

So the sum is: a few hundred pounds and a couple of days to set up (or ten minutes if your accounting tool already does it), against 26 hours a year of your time and thousands in freed-up cash flow. This is one of the clearest returns in all of small-business automation, which is exactly why it's the first thing worth doing.

Common mistakes

Automating this badly is worse than not automating it, so a few traps to avoid.

1. No stop condition. The cardinal sin: chasing an invoice that's already been paid. It happens when the automation reads a stale "paid" status, and it's mortifying — nothing torches trust faster than a firm reminder for money the client sent last week. The paid-status check must run immediately before every send, not once a day.
2. Too-aggressive tone. Some templates read like a solicitor's letter on day one. Match the tone to the rung. Day-0 firmness makes you look like you don't trust your own clients.
3. No human escalation. A ladder with no top rung just nags forever. If day 30 doesn't hand off to a person, you've automated the nagging and kept none of the judgement.
4. Over-engineering. You do not need an AI agent that "understands client sentiment" to send four emails on a schedule. Adding cleverness to a job that runs on three dates and a paid flag just adds cost and things to break. The boring version is the correct version.

What "good" looks like once it's running

Here's the strange thing about a well-built chase system: when it's working, you forget it exists. That's the point. No dashboard you check, no weekly ritual, no shower-thought about the Henderson invoice.

Invoices go out. The day-0 nudge lands and a third of clients pay without you lifting a finger. The seven- and fourteen-day reminders go out in your voice, at the right tone, only to the people who actually owe you. Paid invoices drop out silently. And the two or three genuinely messy ones each month — the dispute, the payment plan, the client gone quiet — land on your desk clearly flagged, so your attention goes to the handful of things that truly need a person.

Your cash comes in faster, your relationships stay warm because the chases are consistent rather than emotional, and the background hum of "who owes me what" goes quiet. It's not glamorous. It's a boring automation doing a boring job perfectly, forever. If you'd like to see this way of thinking in action before spending a penny, our free CRM template is built on exactly the same principle: tidy data, clear rules, and a machine that only bothers you when it should.