Can an AI agent chase unpaid invoices for me?
Can an AI agent chase unpaid invoices for me?
Short answer: yes. In our published cases, an agent checks your open invoices on a schedule, sends a polite reminder before the due date, a firmer one as the invoice ages, and stops the moment payment lands. You set the ladder and the tone. In several cases a person still approves each send. A four-person agency's average days to payment fell from 28 to 14 in six weeks, per its builders. (case)
How the chasing works, from the cases
- It checks every day and flags what is open. One owner set up an AI employee named Marcus on Grok Bot, connected to Stripe and Xero, with its own email address. It runs daily routines to check for outstanding invoices and follow up, reporting back to him rather than sending on its own. The morning it flagged an unpaid invoice from the previous week, he approved the reminder with one click and the client paid the same day, per his post. (case)
- It follows a reminder ladder you set. Solo consultant Lilach Bullock's fixed sequence sends a reminder 3 days before the due date, one on the due date, a firmer one at 7 days overdue, and flags the invoice to her at 14 days. The sequence stops when payment lands. (case)
- It writes each email with the details filled in. A four-person creative agency's pipeline lists the open invoices in FreshBooks; the agent picks a tier by days overdue and writes a personal email with the client name, amount, and link, sends it through Resend, and logs each send in a sheet. The builders wired it in under 2 hours, per the case. (case)
- It gets firmer, then hands the account to a person. A mechanical contractor's follow-up agent checks receipt at day 7, sends a friendly reminder with a pay link at day 20, notes the due date at day 30 and adjusts tone to the payer's history, escalates and flags the account for a human at day 37, sends a final notice at day 45, and hands over to a person at day 50 with the full log. After three late payments it flags the account and suggests deposit terms. (case)
- You can keep approvals in the loop. Solo operator Michael Shaimerden's agent chases quiet proposals and unpaid invoices, and every external action passes one approval queue; money and mass sends never go on auto-send, per the case. (case)
- It can sit inside a wider set of systems. In one three-employee service business, automatic invoices and staged payment reminders were one of five plain systems the owner set up, on a stack costing under $100 a month in total, per the owner; the result for that whole stack is in the table below, not a figure for the reminders alone. (case)
- The tone can still miss. In the agency case, one email came out too casual, so the team tightened the prompt for the oldest tier. Read the drafts for your oldest invoices before you let the ladder run, per that case's own fix. (case)
What the sources report it achieved
Each outcome below is quoted as its case reports it, attributed to whoever reported it: an owner, a builder, a consultant, or a vendor. None of these figures are our measurements, and they are not independently verified by us.
| Case | Outcome, as reported | Source |
|---|---|---|
| Unpaid invoice flagged, paid same day | That morning the agent flagged an unpaid invoice from the previous week; he approved sending a reminder with one click, and the client paid the same day, per his post. | Instagram (@jonathonchilliperez) |
| Onboarding paperwork: 40 minutes to five | Onboarding paperwork fell from 40 minutes to under 5 minutes of her attention. Her monthly report fell from 90 minutes to build to 10 minutes to read. She says late payments drop when reminders stay consistent. Her figures, from her article. | Lilach Bullock |
| Old receivables: $120K down to $45K | In 90 days, bills over 60 days old fell from $120K to $45K. Average collection time fell from 58 days to 34 days. The owner calls it a collections person who never forgets to follow up (vendor's figures; the client is not named). | Enterprise DNA blog |
| Payment collection: 45 days to 12 | After 90 days he reports weekly hours down from 70 to 45, response time from 4 hours to 2 minutes, and payment collection from 45 days to 12 days, with late payments down 70%. Lead conversion rose from 2% to 15% and reviews from 2 to over 10 a month, per the owner. | dev.to (Ken Deng) |
| 106 calls booked, $30,000 signed | In eight months, 38 people booked 106 calls through the system. About $30,000 of work was signed. He still closed the deals himself. Last month, 396 of 845 shipped items went out with no touch from him. Per 100 cold requests, about 40 accept, about 18 reply, and about one becomes a meeting. | Medium |
| Days to pay: 28 to 14 | After 6 weeks, average days to payment fell from 28 to 14. The team missed zero follow-ups. Founder time on invoice chasing fell from about 2 hours a week to zero. One email came out too casual, so the team tightened the prompt for the oldest tier. | DEV Community |
The same job at a larger firm
This case is a larger company, not a small shop, and we do not present its figures as small-business results. It shows the same chasing job with more context wired in: the agent reads each invoice's payment history, contract terms, emails, and disputes, sorts it into a class, and acts inside finance policy; write-offs, credit memos, and legal steps stay with human approval.
| Case | Outcome, as reported | Source |
|---|---|---|
| Aged receivables: 40% to 8% | In the first month, cash on hand grew 122%, per the vendor. The share of receivables that was aged fell from 40% to 8%. The vendor reports these figures. Write-offs, credit memos, and legal steps stay with human approval in the same system. | SourceForge |
More direct answers: all Answers · Every case that reports a money figure: Money found · The numbers behind all our cases: Stats
5 use cases every morning
Get 5 like these every morning, each with its source and a prompt you can copy. One email, 6:30am.