Using AI for Invoice Follow-Up When You Run the Business Alone
By Stacey Tallitsch | May 21, 2026
If you handle the books yourself, you already know the bind. You sent the invoice. The customer agreed to the price. The work is done. Thirty days later, the money is not in your account, and you do not have time to chase it.
The numbers say you are not alone. According to the Intuit QuickBooks 2025 Small Business Late Payments Report, 56% of small businesses are currently owed money on unpaid invoices, with the average exposure sitting at about $17,500. Average payment delays in many sectors now run more than 60 days. Trades operators get the worst of it, with a 67-day average collection period.
The work of getting paid faster used to require an office manager or a part-time bookkeeper. That is still a fine answer if you can afford one. If you cannot, AI tools have closed enough of the gap that a one-person operation can now run a tighter follow-up process than it could two years ago.
Here is how to set that up.
What the AI is actually doing here
The phrase "AI for invoice follow-up" sounds bigger than the work is. Two things are happening under the hood, and you do not need to understand either one in depth.
First, your accounting software watches which invoices are unpaid and how late they are. This is not new. Software has done that for years. What is new is that the system can now draft a follow-up email that reads like something a human wrote, customized to the customer, the invoice amount, and how overdue the bill is. The same idea applies if you draft the email yourself in a tool like ChatGPT or Claude. The AI writes the words. You check them.
Second, the system can decide when to send the reminder. Three days before due. Seven days after due. Twenty-one days after due, with a different tone than the seven-day note. You set the schedule once. The software runs it on every invoice from then on.
Neither of these is magic. They are time-savers. The point is that the writing and the timing happen without you doing them by hand, one customer at a time, at 9 pm on a Sunday.
The two ways to set this up
There are two real paths. Your choice depends on what accounting software you already use.
Path one: Use the AI features already built into your accounting software. QuickBooks Online and Xero, the two most common platforms for small businesses, both ship invoice-reminder features that now use AI to draft the actual message. You turn on automatic reminders, set the days-before and days-after schedule, and the software handles the rest. You can preview every email before it goes out, or let it run on its own once you trust the tone.
Path two: Use a general-purpose AI tool to write the reminders, then send them yourself. This is the cheaper path. It is also the better path if your accounting software does not offer AI drafting or you do not like what it produces. You give the AI a short summary (customer name, invoice number, amount, due date, how late) and a template you wrote once. It produces the email. You paste it into your sent invoices folder and send. The whole thing takes 90 seconds per invoice instead of 10 minutes of you staring at a blank reply box.
For most owner-operators, path one is the right starting point. It does not require any new subscription if you are already on a paid accounting plan. Path two is the right move if you want more control over the language, or if your accounting software is not catching up yet.
A four-step setup
Whichever path you pick, the setup is the same in shape.
Step one: Decide your follow-up cadence. Most small businesses send too few reminders, or send them all in the same tone. A reasonable cadence is a gentle nudge three days before the invoice is due, a firmer reminder three days after due, a more direct note at 14 days late, and a final notice at 30 days late before you escalate. Write that schedule down. Do not improvise it.
Step two: Write your tone for each stage. The three-day-before email is friendly and assumes the customer just needs the date. The 14-day-late email is direct but not hostile. The 30-day-late email is short and serious. If you are using path one, your accounting software handles the tone variation automatically based on how overdue the invoice is. If you are using path two, you give the AI your tone guidelines as part of the template. Either way, you set the tone once.
Step three: Turn the system on and watch the first 10 reminders go out. Do not set it up and walk away. Read every email the system sends for the first two weeks. The drafts will not be perfect. Some will sound off. Some will use language you would not use with a particular customer. Fix those before they go out, and feed the fixes back into your template or your accounting software's settings. After two weeks, the drafts will be close enough to your voice that you can let most of them run on their own. This is the same approach that works for client email drafts in an accounting practice and proposal drafts in an agency. Read the first batch carefully, fix what is off, and let the rest run.
Step four: Decide your manual-intervention rules. Not every invoice should run on autopilot. Your largest accounts should get a personal call from you, not a software email, when they go past due. Pick a dollar threshold (commonly the top 10% to 20% of your accounts) and exclude those from the automated reminders. Your software lets you do this in customer settings. Use it.
What can go wrong
Three things to watch for.
The AI-drafted email can sound generic if you do not give the software enough context. A reminder that says "Hi there, just a friendly note about your invoice" is worse than the template you wrote yourself five years ago. Take the time to set the personalization fields and a tone guide. If the output still sounds canned, that is feedback. Adjust the inputs.
The reminders can damage a customer relationship if a payment crossed in the mail and the software sent a "you are late" email anyway. Always run a one-business-day grace period before the late reminder kicks in. The setting is usually there. Use it.
The system can also feel like progress when it is just busy work. If you are sending more reminders but your cash position is not improving, the bottleneck is not the reminders. It is your payment terms, your credit screening, or your customer mix. AI does not fix any of those. For the broader reality check on what AI productivity tools actually move and what they do not, see the 10x productivity claim, examined.
What to expect after you turn it on
For most small businesses that go from no system to a basic AI-drafted follow-up cadence, two things happen in the first 60 days.
Average days outstanding usually shrink by a week to 10 days. Not because the customers respect AI more than they respect you, but because the reminders go out on time, every time. Most late payments are administrative drift on the customer's end, and a well-timed nudge is enough.
Your weekly time spent on accounts receivable drops by about an hour or two. That is real time, especially for an owner who was doing this work on Sunday nights.
What does not change: customers who were going to pay slow will still pay slow. Customers who were going to dispute will still dispute. AI invoice follow-up shortens the timeline on the customers who were always going to pay. It does not turn bad accounts into good ones.
The cost is small. If you are already on QuickBooks Online or Xero on a paid plan, the AI reminders are usually included. If you are running path two with ChatGPT or Claude, the typical owner-operator spends about $20 a month at the consumer plan tier, which is fine for this kind of work.
Start with the cadence and the tone. Turn on the system. Check the first 10 emails it sends. Adjust. After that, get it off your weekly plate and back to the work that actually pays.
-- Stacey | The Standalone
About the Author
Stacey Tallitsch runs The Standalone, an AI Implementation Diagnostic practice for small business owners. He has 30 years of experience in technology and has written 21 books on systems thinking and decision-making. More than 30,000 students have learned from his online courses.
- Stacey Tallitsch, The Standalone