Stop Chasing Late Payments. Build the Agent That Does It While You Sleep.
To automate invoice follow-up for a small business, connect your invoice list (your accounting software, a spreadsheet, or a CRM) to a no-code workflow tool like n8n, Zapier, or Make. Add a daily scheduled check that flags anything unpaid past 30 days, then let an escalating sequence, a friendly nudge on day one, a firmer message on day seven, a final notice on day fourteen, go out automatically by email, SMS, or WhatsApp until the client pays or a human steps in.
You did the work. You sent the invoice. Then you became a debt collector nobody warned you about.
It is 9 p.m. On a Thursday. Marisol runs a five-person landscaping and hardscape crew outside Austin, Texas. The last truck is back in the yard, the last client photo is posted, and she is sitting at her kitchen table typing the same email she typed last month.
"Hi James, just a friendly reminder that invoice #4471 for the Okafor patio job is still outstanding..."
She has sent some version of that sentence four times this week. Different names. Same words. Same low simmer of a business owner who finished the job weeks ago and is still waiting to get paid for it.
The math you don't see until it's this bad
Marisol is not unusual. She is the median.
In the UK it is worse. Sage's SME Pulse for mid-2026 put it plainly: nearly half of all small-business invoices, 49 percent, sit overdue at any given moment, and late payments now cost the UK economy an estimated £11 billion a year in stalled cash flow. Different currency, same story: the job is done, the invoice is sent, and the money is somewhere between "processing" and "forgotten."
Why you keep doing this by hand
Every month it is the same loop. Pull the aging report. Decide who to chase. Rewrite the same email. Log it. Repeat.
None of that grows the business. It is not client work. It is not sales. It is unpaid admin sitting on top of work you already finished, wearing the disguise of being a considerate business owner.
You are not being nice by handling this manually. You are being the bottleneck between the work you already did and the money you already earned.
What is an invoice follow-up agent?
A chatbot answers the question "when is my invoice due." An agent does not wait to be asked. It checks your invoice list every morning on its own, decides which ones crossed a line you set, drafts the right message for where that invoice is in its life, sends it through the channel your client actually reads, and escalates on a schedule you never have to remember, because it is not relying on your memory in the first place.
That is the whole difference between automation and an agent: automation waits for a trigger you pull. An agent watches, decides, and acts.
"Isn't this just an email reminder with a delay timer?"
A basic reminder is one email, sent once, the same for every client. An agent is a ladder. It treats a client who is one day late differently from a client who is two weeks late, and it treats a first-time miss differently from a repeat offender if you let it look at payment history. A delay timer cannot tell the difference. A well-built agent can, and that difference is what keeps your reminders from sounding like a robot repeating itself.
How do I automate invoice follow-up for my small business?
You do not need to be a developer. You need a workflow tool, your existing invoice data, and about an afternoon. Here is the build, step by step.
- Put your invoices somewhere a workflow tool can read them. This can be QuickBooks, Xero, Wave, or something as simple as a Google Sheet with columns for client, amount, due date, and status. The tool does not care which; it just needs one place to check.
- Add a daily scheduled trigger. In n8n, Zapier, or Make, this is a built-in node that fires once a day, for example every morning at 9 a.m. No one has to remember to run it.
- Pull the unpaid invoices and check their age. A filter or "if" step compares today's date to the due date and flags anything unpaid past your threshold, most builders use 30 days, but a service business that invoices weekly might set it at 7 or 14.
- Branch into an escalation ladder. Day 1 gets a friendly nudge. Day 7 gets a firmer message. Day 14 gets a final notice, and can trigger a task for you to make an actual phone call instead of sending a fifth email into the void.
- Add a judgment layer instead of a static template. This is what turns a workflow into an agent. Instead of a static message, send the invoice details and the client's payment history to a language model (a Claude or GPT call is a single node in most of these tools) and ask it to draft the reminder in your voice, and to flag if this client has missed twice before and needs a call, not another email.
- Send through the channel your client actually reads. Email still works for most B2B clients. For consumer-facing service businesses, plumbing, salons, contractors, a text through Twilio or a WhatsApp Business API message gets opened faster than an email that gets buried.
- Log every action back to your sheet or CRM. Every message sent, every reply received, every payment logged, so you can see the whole trail without digging through your inbox.

What Marisol's agent actually does now
Marisol rebuilt her Thursday-night routine as a single n8n workflow over one weekend. It checks her invoice sheet every morning. Anything past 30 days gets a day-1 email drafted in her tone, referencing the actual job. Anything past 37 days gets a firmer message. Anything past 44 days gets a final notice and adds a task to her own calendar: call this client.
Two weeks after she turned it on, the Okafor patio invoice paid after the second reminder. A larger commercial client paid on the very first one, before she would have even remembered to chase it. She still makes the occasional phone call for the hard cases. She has just stopped being the one who has to notice they are hard cases in the first place.
| Chasing it yourself | The agent chasing it | |
|---|---|---|
| Time per week | 6 to 10 hours: aging report, drafting, logging | 10 to 20 minutes reviewing exceptions |
| Consistency | Depends how busy or tired you are | Same ladder, every invoice, every time |
| Tone | Ranges from too soft to accidentally sharp | Drafted in your voice, adjusted to payment history |
| What happens on vacation | Nothing. Invoices sit. | Nothing changes. It keeps running. |
The trend is bigger than one workflow
This is not a one-off idea. Builders on X are shipping single-purpose versions of exactly this agent right now: one called Tallywing chases open invoices across email, SMS, WhatsApp, and voice, and hands the founder a one-page weekly memo instead of ten hours of manual chasing. The pattern shows up because the pain is universal and the fix does not require a finance team, only a workflow tool and an afternoon.
What you have after this
A week in, you notice the emails you used to write yourself are already sent, and already sound like you.
A month in, you stop opening your invoice sheet out of anxiety and start opening it to see what got paid.
A quarter in, your average days-to-pay has dropped enough that you notice it in your bank balance before you notice it in a report.
Same clients. Same jobs. Same invoices. A completely different relationship with your own money.
Is it legal to text or WhatsApp customers about unpaid invoices?
In most places, yes, if the customer agreed to be contacted that way, which is why it is worth adding a line to your invoice terms or intake form saying you may follow up by email, text, or WhatsApp about payment. Rules on unsolicited texts vary by country and by whether the message is transactional versus marketing, so check your local consumer-messaging rules before turning on SMS or WhatsApp reminders, and always give an easy way to reply or opt out.
What if a customer disputes the invoice instead of paying it?
A well-built agent should stop the escalation ladder the moment a client replies, not keep sending firmer messages into an active conversation. Route any reply straight to a human inbox or task, and only let the automated sequence resume for invoices where there has been no response at all.
Do I need to know how to code to build this?
No. N8n, Zapier, and Make are visual, drag-and-drop workflow builders. The most technical part is writing a simple condition like "days overdue is greater than 30," which most of these tools let you build by clicking dropdowns rather than writing code.
How much does an invoice follow-up agent cost to run?
N8n can be self-hosted for free or run on a cloud plan starting around $20 to $24 a month. Zapier and Make have free tiers that can handle a small volume of invoices. Add a few cents per message if you use SMS or WhatsApp through a provider like Twilio, and a small per-call cost if you add an AI drafting step. For most small businesses this runs a few dollars a month, far less than the hours it replaces.
Find your first high-payback workflow.
See the SprintSources
- Xero: New Xero Data Shows U.S. Small Businesses Face Mounting Pressure as Payment Delays Rise
- Sage: UK SME profits hit a four-year high, with faster payments key to unlocking further growth
- Xero Blog: Small businesses getting paid quicker, but still late
- n8n Docs: Schedule Trigger node
- n8n Docs: If node
- YouTube: Stop Chasing Late Invoices, Build an Auto Debt-Recovery Engine in n8n (Irvan Hardiansyah)
- X post: Tallywing, an always-on AI agent that chases open invoices (@polsia)
Find your first high-payback workflow.
Book a free conversation or start with the fixed-fee Sprint.
Keep reading
Your Biggest Decision This Quarter Doesn't Need a CFO. It Needs Three AI Agents.
To automate a big, one-off business decision, such as opening a new location, hiring a manager, or taking on debt, build a small team of coordinating AI…
Back-office AIYour Inventory Doesn't Need a Spreadsheet. It Needs an Agent.
To automate inventory management with an AI agent, put your stock in one shared sheet, connect it to a no-code tool like n8n, add a Claude-powered agent…
AI Tools & SkillsHow to Install an AI Skill for Customer Welcome Emails
Yes. A ready-made AI "skill", a small instruction file that teaches an AI agent to do one job well, can now write and set up a full customer welcome email…