Every business we work with has someone who opens supplier invoices, checks them, codes them and files them. Ask what that costs and the estimate is usually too low, because the visible part of the task is only the data entry.

The expensive part is verification: was this the rate we agreed, was this freight charge correct, did we actually receive all of it.

What actually gets automated

Capture is largely solved. A bill arrives, header and line data is read, a draft is created in Xero against the right supplier. That removes re-keying and is worth having on its own.

What is rarely in place is the layer above it: comparing each line against the price you agreed, checking received quantities against the purchase order, and allocating landed cost so margin stays honest.

Where the accuracy risk sits

Field capture accuracy is not the real risk. The real risk is silent failure: an invoice that reads cleanly, posts cleanly, and is wrong. A five percent rate discrepancy on a high-volume line will not look unusual on a single bill and will not be caught by anyone reading quickly.

So the system has to be built to argue with the invoice, not only to read it. Every line is compared against a reference, and anything outside tolerance is surfaced with both numbers side by side.

What should not be automated

Approval and dispute. A system can assemble the case: here is the agreed rate, here is the billed rate, here is the delivery record, here is the variance. A person decides whether to pay, query or hold. That boundary is what keeps the finance team in control rather than supervising something they do not trust.

A realistic sequence

Start with capture and coding for your top ten suppliers by volume. Add rate matching next, because that is where the recovered money is. Add landed cost allocation last, once the data feeding it is reliable.

Done in that order each stage pays for the next, and nobody is asked to trust a system they have not yet seen behave.