New AI Tool Predicts If Your Customers Will Pay You on Time

Waiting to get paid is one of the biggest headaches for any business. A new research paper from computer scientists offers a possible fix: an AI-style tool that scores how likely a customer is to pay you — and roughly when. Called Compensability Capacity Assessment (CCA), it works by analyzing past invoice histories between a buyer and supplier, similar to how a credit score predicts whether you'll repay a loan.

The tool was tested on nearly 750,000 real invoices from 2012 to 2023. The results are striking: in both test scenarios, 89–91% of expected payments arrived within 7 days of the invoice being issued, and 94–96% within 30 days. That means payment timing is much more predictable than many businesses assume. The highest-scoring customer relationships recovered 92.5% of the best possible payment relief, while low-scoring ones were far riskier — only about 58% showed a strong chance of getting paid within a month.

Why does this matter? For a small supplier, choosing which contracts to accept — or which late-paying client to drop — is a survival decision. With a tool like this, a company could rank potential partners the way lenders rank borrowers. It could also help businesses negotiate better terms or plan when to expect cash, reducing the need for emergency loans or stressful scrambling to cover payroll.

The catch: the model is good at ranking who is likely to pay sooner, but weaker at predicting the exact day money will land. And like all forecasts, it works on historical patterns — it won't catch a customer suddenly going broke. Still, this research points to a future where 'will they pay?' is no longer a guess, but a data-driven answer.