Automating invoices is a decision with three quite different answers, and the right one depends on a measurement most organisations skip. A week of recording where time and elapsed days go points clearly at one of three fixes, and the three are not substitutes for each other. The measurement costs an hour a day and frequently redirects the project entirely.
Keying, waiting, or exceptions
Time ten invoices from arrival to approval and record the minutes and the elapsed days separately. Then count last month's exceptions by reason. Keying minutes point at capture. A large gap between minutes and days points at approval routing. Exceptions dominated by missing orders and receipts point upstream, at a free fix.
What each fix looks like
Capture removes typing and transcription errors. Workflow routes to named approvers with cover and escalation, attacking the waiting. Upstream discipline means a purchase-order threshold people know and receipts recorded at the delivery point, which raises the ceiling on everything else and costs nothing but persistence.
Why the order matters
Buying capture when the constraint is waiting produces accurate data that still sits for a week. Buying matching when orders are not raised produces software with nothing to compare. Doing the upstream work first raises the ceiling for whatever you buy afterwards, which is why it is worth doing even when software is definitely coming.
Questions people ask about automate invoices
Can we skip the measurement?
You can, and roughly half the time you will buy the wrong thing first. An hour a day for a week is cheap insurance.
What if the answer is all three?
Sequence them: upstream discipline, workflow, then capture. Each makes the next more valuable.
Does size change the answer?
Volume changes the size of the prize, not which step is the constraint. Small teams have the same three failure modes.