Accounts Payable Automation

What AP automation actually does

The accounting software already handles what most people associate with accounts payable - bill creation, supplier records, payment runs. The gap isn't the accounting system. It's everything between receiving a supplier invoice and posting it to the ledger: the coding decisions, the validation checks, the approval routing, and the exception handling. That's where most AP labour sits, and that's what automation addresses.

For a 20-person Australian business processing 80 invoices a month, staff time per manually processed invoice typically runs 20 to 30 minutes - data entry, coding, chasing approvals by email, resolving mismatches, and occasionally unwinding a duplicate that cleared. The ATO's cost benchmark of AU$27.67 per manually processed invoice translates to AU$2,214 per month in direct processing labour at that volume. Full AP automation doesn't make that number zero, but it routinely reduces it by 60 to 70% - by shifting the AP officer from processing every invoice to reviewing only the flagged exceptions.

Why Xero and MYOB aren't the full workflow

Both Xero and MYOB handle the ledger side of accounts payable competently. They record bills, apply tax codes, manage supplier contacts, and support basic approval routing. What they don't do is sit upstream of their own data entry and validate what's coming in.

When an invoice arrives in Xero's email-to-bills inbox, it extracts the data and creates a draft bill. If the bank account number on that invoice differs from the one stored for the supplier, Xero has no mechanism to flag that discrepancy - it captures what it receives. If the same invoice was submitted two weeks ago with a slightly different reference number, the duplicate won't be reliably caught. If the ABN on the invoice doesn't match the entity in the ATO's register, the bill gets created regardless.

These aren't criticisms of Xero or MYOB. They're accounting systems, not AP controls systems. The control layer belongs upstream - between invoice receipt and ledger entry - and needs to be purpose-built for that function.

What a complete workflow looks like in practice

Invoices arrive at a single intake point regardless of how they're sent - emailed to a processing address, uploaded directly, or received through a supplier portal. Everything enters the same queue.

Before a human sees the invoice, the controls layer runs. The supplier's bank account number is compared against the stored record - a mismatch holds the invoice immediately, with a specific flag explaining what changed, before it can reach any approval queue. The invoice is checked for duplicates against reference number, amount, and date - not just invoice number alone, because reference numbers get reused. The ABN is verified against the ATO's ABR. If a purchase order exists, invoice lines are matched against it at line level, not header level.

Account coding happens next, at the line level, based on that supplier's full coding history. The same freight supplier gets coded to the same accounts every time - not because someone remembered, but because the system has learned the pattern from every prior invoice. GST is applied per line, which matters for mixed invoices where not every line carries the same treatment.

By the time the invoice reaches an approver, it's been through all of that. The approver sees a clean, validated invoice with confirmed bank details, verified ABN, no duplicate flag, and line-level coding that reflects prior practice. After approval, the bill publishes directly to Xero or MYOB with everything applied: account codes, tax codes, cost centre allocations, entity assignment. No re-entry.

Where most implementations fall short

The majority of Australian businesses that have implemented invoice automation have the first and last steps - capture and ledger sync - but not the middle. OCR tools extract data from invoices and push them to Xero with account code suggestions. That's real productivity savings. It's not a controls layer.

The distinction matters because the controls layer is what prevents the losses. A duplicate invoice that was paid and not recovered looks like a supplier payment in the accounts until someone notices and queries it. A bank detail fraud that succeeded looks like a regular EFT transaction until the real supplier chases the missed payment. Neither of these shows up in the processing-time calculation.

The businesses that get hit by payment redirection fraud - which cost Australian businesses AU$152.6 million in 2024 according to the ACCC - almost always have OCR running. They've automated the data entry. They haven't automated the check that would have caught the changed bank account number before the payment was approved. Full AP automation isn't a faster version of the current process. It's a different process - one where the controls that used to depend on individual attention run automatically on every invoice, every time.