How to use AI to create a simple invoice follow-up system that chases late payments without you lifting a finger

You are owed money right now, and you are spending hours chasing it manually. This post walks you through building a 5-step automated invoice follow-up system that sends personalized emails without you touching it. Businesses that automate payment reminders get paid up to 3x faster than those doing it by hand — that is not a promise. That is what the data from FreshBooks{:target="_blank"} shows.

According to QuickBooks{:target="_blank"} data, small businesses carry an average of $84,000 in outstanding receivables at any given time. A 2023 Xero study{:target="_blank"} found 48% of invoices are paid late. Chasing those invoices manually costs a solo operator about 14 hours per month — time you are not getting back.

---

What you need before you start

Make{:target="_blank"}: Visual automation platform (called "Scenarios" for workflows, "Modules" for steps — keep that terminology in mind). Free tier covers 1,000 operations per month, which handles 20–50 invoices monthly without issue. Paid plans start around $9/month as of publication.

Your invoicing app: This system works cleanly with QuickBooks Online{:target="_blank"}, FreshBooks{:target="_blank"}, or Xero{:target="_blank"}. All three have native Make integrations with invoice-status triggers. Wave{:target="_blank"} also works. If you use PayPal, Stripe, or Square invoicing, you will need a webhook workaround — that is a separate build.

Gmail or Outlook (Microsoft 365): Your emails send from your actual address, not a generic notification address. This matters for open rates and deliverability. Connect whichever you use.

An AI model inside Make: Make has a native OpenAI module for GPT-4o{:target="_blank"} and supports Anthropic Claude 3.5 Sonnet{:target="_blank"} via its HTTP module. GPT-4o follows structured output instructions reliably. Claude 3.5 Sonnet produces a slightly warmer tone. Either works here.

Time required: 2–3 hours for a non-technical user building this for the first time. One-time build that runs indefinitely after that.

Skill level: No coding required. Basic comfort navigating app dashboards helps.

---

Build your 5-step automated invoice follow-up sequence

The sequence uses five touch points, each with a distinct tone. This cadence is what AR automation platforms test and settle on:

• Day -3: Heads-up reminder before the due date
• Day 0: Due date reminder
• Day +3: Polite nudge
• Day +7: Firmer notice
• Day +14+: Escalation message

Step 1: Create a new Scenario in Make

1. Open Make{:target="_blank"} and log in.
2. Click Create a new scenario in the top right corner. You should see a blank canvas with a prompt to add your first Module.

Step 2: Set your invoice trigger

3. Click the + icon and search for your invoicing app (QuickBooks Online, FreshBooks, or Xero).
4. Select the trigger Watch Invoices or Invoice Status Changed — the exact label varies by platform. You should see a connection prompt asking you to authorize your invoicing account.
5. Connect your account and set the trigger to fire when an invoice status changes to Overdue or when a due date passes without payment. You should see a confirmation that the trigger is active.

This trigger is the starting point for everything. Every branch of your sequence flows from this single event.

Step 3: Add a Router module to branch by days overdue

6. Add a Router module after your trigger. A Router splits your workflow into separate paths based on conditions.
7. Create five paths. Label them: -3 days, Day 0, +3 days, +7 days, +14+ days.
8. Set the filter condition on each path using the invoice due date field. Calculate the difference between the current date and the due date. You should see each branch activate only when that condition is true.

The Router is what gives each message a different tone. Without it, you send the same email every time — which is what QuickBooks' built-in reminder does, and why it underperforms.

Step 4: Add an AI module to write each email

9. On each Router branch, add Make's native OpenAI module (for GPT-4o) or an HTTP module (for Claude 3.5 Sonnet).
10. Connect it to GPT-4o or Claude 3.5 Sonnet using your API key from OpenAI{:target="_blank"} or Anthropic{:target="_blank"}.
11. For each branch, paste a prompt that includes the client's name, invoice number, amount, and due date pulled from your invoicing app. Adjust the tone instruction per branch.

Use this prompt structure for the Day +3 path as your template:

AI prompt — Day +3 polite nudge:

• You are writing a payment reminder email on behalf of a small business owner.
• Tone: warm, polite, non-accusatory. This is a soft reminder.
• Client name: [map from invoicing app]
• Invoice number: [map from invoicing app]
• Amount due: [map from invoicing app]
• Due date: [map from invoicing app]
• Write a subject line that includes the client's first name and invoice number.
• Write the email body in 3–4 sentences. Do not use exclamation marks. Do not threaten.
• Output format: Subject: [subject line] / Body: [email body]

For the Day +7 path, change the tone instruction to: professional, direct, and firm — this is the second follow-up and payment is now one week late.

For the Day +14+ path, add: Reference that continued non-payment may result in late fees as specified in the original invoice terms. Check your original invoice terms first — in the US, late fees must be in your contract to be enforceable.

Personalized subject lines increase open rates by 37% compared to generic ones, per GoCardless{:target="_blank"} A/B testing. The prompt above handles that automatically.

Step 5: Connect Gmail or Outlook to send the email

12. After each AI module, add a Gmail or Microsoft 365 Email module.
13. Map the To field to the client email from your invoicing app.
14. Map the Subject and Body fields to the AI output from Step 4.
15. Activate the Scenario. You should see it listed as On in your Make dashboard.

Run a test with a real invoice in draft or a test account before going live. You should receive the email in your own inbox with populated client data.

---

When something goes wrong

The trigger fires but no email sends. The AI module likely returned an error from the API. Open the Scenario execution log in Make and look for the HTTP module status. A 401 error means your API key is wrong or expired. Re-enter the key in your connection settings.

The email sends but client data is missing — the name or invoice number shows as blank. The data mapping broke between your invoicing app module and the AI prompt. Open the AI module, click into the prompt field, and re-map each variable by clicking the field name from the data panel on the right. Make can lose mappings if you reconnect an integration.

Emails are going to spam for some clients. You are sending from a domain that has no automation history. Sending through a connected Gmail or Outlook account reduces this significantly — confirm your email module is using an OAuth connection, not raw SMTP credentials. Also check that your email account has two-factor authentication enabled, as some providers flag unverified sending.

---

What to do next

After your sequence runs all five steps and a client still has not paid, the automation's job is done. At that point you have a documented paper trail of five contacts — useful if you escalate to collections or small claims. Set a Make Scenario to flag that invoice in your invoicing app as requiring manual review at day 21. That keeps your inbox out of it until a human decision is actually needed.

Read [how to set up a client onboarding automation in Make](INTERNAL_LINK: client onboarding workflow automation Make small business) to extend the same system logic to the front end of your client relationships.

---

FAQ

Can I use Zapier instead of Make for this system? Yes. Zapier{:target="_blank"} connects to QuickBooks Online, FreshBooks, and Xero natively. Zapier's late-2024 "Zap Steps with AI" feature lets you use GPT-4o or Claude 3.5 Sonnet inside a Zap without a separate API key, which lowers setup complexity. The free tier covers 100 tasks per month — workable for smaller invoice volumes, but Make's free tier is more generous at 1,000 operations.

What if I don't want to build this myself? Chaser{:target="_blank"} is a dedicated accounts receivable chasing platform that handles the entire sequence natively, including tone variation and escalation logic. Plans start at $50/month. If your time is worth more than the build cost and you send a high volume of invoices, Chaser is the cleaner option. The DIY Make build costs nothing beyond your API usage once it is set up.

Will clients find automated reminders impersonal? The AI personalization in this system addresses that directly. Emails include the client's name, the specific invoice number, and vary in tone based on relationship context if you add that to the prompt. A generic "Invoice Overdue" blast from accounting software is impersonal. An email from your actual address, referencing the right invoice, written in a human tone, is not obviously automated.

Do I need to tell clients I'm using automated reminders? No legal requirement exists in the US or UK to disclose that a reminder email was generated by automation. The email comes from you, references a real invoice, and requests legitimate payment. Standard invoice terms and your original agreement govern the relationship — not the sending mechanism.

What's the difference between this and QuickBooks' built-in reminders? QuickBooks Online's built-in Reminders feature sends a single static message. It does not vary tone across multiple touchpoints, does not personalize content per client, and does not support escalation logic. The system in this post does all three. FreshBooks has similar limitations on its native reminder scheduling.