Case study ยท Logistics

Invoice-to-PO matching: from 3 days a week to automatic

The problem

A 60-person logistics company was spending three days a week manually matching freight invoices to purchase orders across four different systems. The process ran on spreadsheets, tribal knowledge, and one person who knew where everything lived.

What we found

When we went inside, we discovered the matching logic was actually consistent โ€” the same rules, every time. The only reason a human was doing it was because nobody had ever mapped it out end to end.

What we built

A custom AI agent that handles invoice-to-PO matching automatically, flags exceptions for human review, and generates a daily exceptions report. It took six weeks from discovery to live.

Result: three days a week of manual matching became automatic, with humans only touching the exceptions. Live in 6 weeks.

This is how we work: we go in, learn how the business really runs, find what AI can fix, and build it.