Overdue invoice collection: how to automate it without chasing paid invoices

Where does overdue invoice collection start?

Start with the data, not the message. A polite reminder sent for the wrong invoice causes more damage than a late collection attempt.

The basic flow has an invoice source, a payment source, a contact policy, and a record of every interaction. If those four points aren’t clear, the system can chase customers who already paid or escalate cases that are in dispute.

In Colombia, collection must follow commercial and data rules. Colombia’s Law 2024 of 2020 regulates payments caused by commercial acts, with exclusions. It also sets, from its second year in force, a maximum term of 45 calendar days for many commercial obligations. If there is late payment, it allows recovery of proven collection costs, in addition to rules on default interest. This isn’t legal advice; validate your case with your legal team.

There’s also a practical reason to organize receivables. The Atradius Payment Practices Barometer 2026 reports that in North America, companies make an average of 43% of their B2B sales on credit terms, and that late payments affect seven out of ten surveyed companies. Many delays come from customer liquidity, but also from internal approvals and banking friction.

That’s why we measure before we automate. We review days sales outstanding, overdue amount by aging bucket, kept payment promises, responses received, and claim errors. Then we compare the new flow against that baseline. If a metric doesn’t improve, automation only moved the work somewhere else.

What data do you need before automating reminders?

For companies in Colombia, if you process personal data in collections, you need authorization and a clear purpose. Colombia’s Law 1581 of 2012 requires prior and informed authorization, with exceptions, and requires security measures to prevent unauthorized use or access.

• Customer: legal name, NIT or tax ID, authorized contact, allowed channel, and account owner.
• Invoice: number, issue date, due date, amount, currency, purchase order, and proof of delivery.
• Status: issued, sent, accepted, overdue, paid, partially paid, disputed, or written off.
• Payment: bank, date, amount, reference, withholding, discount, credit note, and reconciliation against invoice.
• Collection activity: last message, customer response, payment promise, next action, and case owner.
• Risk: strategic customer, late-payment history, open claims, high amount, or legal alert.

How do you automate overdue invoice collection without chasing paid invoices?

A good flow separates simple rules, reconciliation, and human judgment. Not everything should go out by email or WhatsApp on the same day.

1. 1. Normalize invoice statuses. Define one status per invoice and one rolled-up status per customer. If a customer has five invoices, the message should talk about the real balance, not each document in isolation.
2. 2. Reconcile before contacting. Cross-check bank payments, cash receipts, credit notes, and withholdings. If this part fails, review a bank reconciliation automation before sending bulk reminders.
3. 3. Create contact rules. Example: notice before the due date, reminder on the due date, second reminder with support detail, and a human task if there’s no response. Adjust by customer type and amount.
4. 4. Record every event. Store the message sent, channel, response, payment promise, and blocker. Microsoft documents in Dynamics 365 Finance that tracking steps avoids repeating communications from a stage already completed, and that the oldest overdue invoice can determine the customer’s collection stage.
5. 5. Escalate exceptions. A person should review disputes, strategic customers, unreconciled payments, high amounts, and cases with legal risk. Automation prepares the file; the decision shouldn’t always be automatic.
6. 6. Measure before and after. Compare DSO, overdue amount, response rate, kept promises, claim errors, and time to close. At Quarl, we design these metrics before building the flow.

What does an accounts receivable follow-up flow look like in a real company?

This is an illustrative example, not a client. It helps ground the decision before buying software or building an integration.

A B2B company sells monthly services. It has electronic invoicing, an ERP, bank statements, email, and WhatsApp Business. The receivables team reviews a spreadsheet every morning and writes messages by hand.

The system takes overdue invoices from the ERP, checks reconciled payments, groups by customer, and excludes disputed invoices. If proof of delivery is missing, it creates a task for operations. If the customer paid partially, it calculates the balance and avoids claiming the full amount.

Then it sends messages according to policy. The first message reminds the customer of the invoice, amount, due date, and payment link. If the customer replies “already paid,” the system asks for proof through an allowed channel and marks the case as pending reconciliation. If the customer replies “I didn’t receive the service,” it sends the case to human review with the full history.

In Colombia, electronic invoices have specific requirements if they are to be treated as negotiable instruments. DIAN lists prior validation, due date, acknowledgment of receipt, receipt to satisfaction, and express or tacit acceptance in the RADIAN FAQ. RADIAN records electronic invoices considered negotiable instruments and allows traceability.

It’s a mistake to confuse technical validation with commercial acceptance. In Concept 9471 of 2026, DIAN stated that the “Approved with notification” status does not replace acknowledgment of receipt, receipt of the good or service, or invoice acceptance.

DIAN also states that there is a web service to check invoice statuses and events. In its frequently asked questions, it explains that tacit acceptance occurs when there is acknowledgment or proof of receipt and three business days pass without rejection. Those events help decide when to insist and when to review support documents.

What software works for accounts receivable follow-up?

The decision depends on where your data lives and how much control you need over the rules.

If you need to connect ERP, bank, CRM, and email, the work is usually in integrations and APIs. AI can come later, inside an AI automation effort, if it improves a concrete metric.

What we do

• AI automation Applied where there is volume and stable rules, not where there is expectation.
• Integrations and APIs MCP servers against the systems already in operation.
• AI for finance Consolidation and analysis with figures traceable to their source.