← All articles

Unpaid invoices: tracking receivables without chasing anyone

The work is done. The invoice, sent. And one ordinary day you realise that customer still has not paid, and that they should have paid two months ago.

In many small companies, money already earned is the worst-watched capital they have. Not because nobody cares, but because the information is scattered: the invoice is in the invoicing program, the agreed due date in an email, the "I'll sort it this week" in a chat message, and what has actually arrived at the bank somewhere else. Nobody has the whole list in front of them, and what is not seen is not chased.

The problem is not chasing. It is knowing whom to chase.

When a small business stops chasing a payment, it is rarely out of neglect. It is because every time they want to, they have to rebuild the situation from scratch: which invoice it was, for how much, when it was due, whether the customer said anything, whether a reminder already went out last week.

That is twenty minutes of work per customer. Multiply it by fifteen and you understand why it keeps getting left for tomorrow. The solution is not more hours: it is for the list to exist and fill itself.

The three steps, in order

1. A single list of due dates. There must be one place where, in ten seconds, you see every outstanding invoice with its due date, amount and days overdue. If you have to open three programs to find out, the system has already failed. This first step can be done even with a spreadsheet, and it already improves things a lot.

2. The list fills itself. This is where AI genuinely adds something. A document-extraction system reads the issued invoice and logs the number, customer, amount, date and due date by itself. If the customer replies by email with a payment date, that gets recorded too. You key nothing: you review.

3. Staged reminders, not a flood of emails. The system reminds five days before the due date — a friendly reminder prevents more late payments than ten claims afterwards — on the due date, and at seven and thirty days overdue. Each one has a different tone, and the thirty-day one should no longer be an automated email: it is an alert for you to step in.

What the automation must never do

This is where many systems do more harm than good. Three rules worth being clear about before building anything:

A concrete example

An installations company with forty recurring customers invoiced well and collected late. The manager spent a whole morning every month reconstructing the picture and writing emails, and even so three or four always slipped through.

They set up the three-step circuit. Every issued invoice is logged by itself with its due date, and every Monday a summary arrives in three blocks: due this week, less than 15 days overdue, and you need to call these. The first two blocks are resolved with reminders the manager approves in five minutes. The third is short and, at last, is the one that actually deserves his time.

The result was no accounting magic: it was the morning recovered and, above all, no longer discovering unpaid invoices by chance.

The part to look at carefully: privacy

This list contains sensitive data: who your customers are, how much they owe you, and who pays badly. Before putting it anywhere, ask three things:

This article is general information, not legal advice for a specific case.

About this article

iaetica.cat is a space for writing in the open: what I am learning about artificial intelligence, its ethics and its law.

— Laura Graells