The problem
Inbound invoices used to mean someone reading each one, keying it in, matching it against the supplier statement, and catching the duplicates by eye. Slow, dull, and exactly the kind of work that gets skipped when everything else is busy — which is precisely when a duplicate slips through and gets paid twice.
This was our own problem, not one we read about in someone else's case study. The same back-office grind every business with suppliers knows. So rather than tell anyone else how to fix it, we fixed ours first.
Our approach
We built a system that reads incoming invoices, reconciles them and the matching supplier statements against Xero, and flags duplicates before anything reaches a payment run. The reading is the part people assume needs a human — pulling the supplier, the amount and the line items off a document every supplier formats differently — and that is the part we automated. The reconciliation against the ledger and the statement follows, so a discrepancy or a double-up surfaces while there is still time to act on it.
We built it in-house, using AI to build it faster than we otherwise could have. A person still makes the call on what to pay — the system does the reading, the matching and the flagging; the decision stays ours.
Because it runs against our own books, the team that built it lives with it. When it gets something wrong, we feel it before any customer would.
The outcome
The system clears 50 to 80 of these jobs a week for us automatically — around ten hours that used to be done by hand, every week, handed back to the team. Duplicates get caught before payment, not chased down afterward. The dull, error-prone part of accounts admin now just happens in the background.
The numbers here are ours: we built this for ourselves, we run it every day, and it earns its place. That is the only claim worth making about AI — not what the technology might do, but what we have already made it do in our own business.
If your team is still reading and keying invoices by hand, that is almost certainly a job we could take off them — the same way we took it off ours.
