November offer: EXTRA 30% off all courses & bundles. Use NOVEMBER30 at checkout. Ends 30 Nov.

AP Automation in 6 Steps: From Invoice Receipt to Payment

All six steps of AP invoice processing can be automated, from receipt to payment, but not every invoice. Here is where to start, with or without e-invoicing.
Oct 24 / Peter Ilas, FCCA
Ask a finance team how much of their AP invoice processing they could automate, and most will guess 50%, maybe 80%. The honest answer is higher than most people expect, with one important condition.

In this article, I go through AP automation step by step: what the system does, what people still do, and why e-invoices make the difference. I also share what I learned when automation did not work as planned.

The short answer

All six steps from invoice receipt to payment can be automated: receive and validate, match, code, approve, post and pay. That is 100% of the steps, not 100% of the invoices; exceptions still go to people. You can start today with PDF invoices and OCR or AI capture. E-invoices take you much further, because they deliver every invoice line as data.

Why AP invoice processing is still so manual

In many AP departments, the process still looks as it did when invoices came on paper. An invoice arrives by email. Someone enters or checks the data, matches it to the purchase order and goods receipt, sends it for approval and chases the approvers. Only then is it posted and paid.

Every step needs a person, and every person adds waiting time. On average, AP teams take 8.2 days to process a single invoice (Ardent Partners, State of ePayables 2025); in a largely manual process, 8–14 days is common. Early payment discounts are lost, and the AP team spends its time typing and chasing instead of analysing.

The 6 steps of AP automation, and what people still do

Full AP automation follows the invoice through six steps. Each step builds on the previous one, so the order matters. People are still involved in the process but in a different way.
StepWhat the system doesWhat people still do
1Receive, record and validateCreates the record; checks supplier, VAT/tax ID, totals and duplicatesClear rejected invoices with the supplier
2Match to the PO or contractRuns a 2-, 3- or 4-way match against the PO, goods receipt and quality checkResolve price and quantity differences
3Code non-PO invoicesProposes the GL account, cost center and other dimensions from vendor and invoice patternsReview new vendors and new types of cost
4Approve low-risk invoicesApproves invoices that meet all the rulesApprove exceptions, high amounts and new vendors
5Post to the ledgerPosts once validation, matching and approval are completeReview the postings and improve the rules
6Pay when duePrepares the payment run on the due dateMonitor the payment run and its controls
In my experience, step 1 alone saves about 15–20% of the AP workload, and steps 1 and 2 together about 40–50%. Steps 3 to 6 take over the rest: coding, approvals, postings and payments.
AP automation in 6 steps from invoice receipt to payment, with the workload saved by steps 1 and 2
Steps 1 and 2 alone save about 40–50% of the AP workload

With or without e-invoices: what really changes

You don't need e-invoicing to start. Many companies begin with PDF invoices and OCR or AI capture. It is a real first step, but it has a limit.

OCR reads a picture of the invoice. In practice, it captures the header: vendor, invoice number, dates, totals, VAT/tax and PO number. In my experience, OCR tools reach about 80% accuracy and completeness, so a person still checks every invoice.

A structured e-invoice is data, not a picture. The system receives every field and every line exactly as the supplier issued it. That makes line-level matching possible, and with it steps 2 to 5 without a person.
PDF invoice with OCR capturing header fields only, next to an e-invoice with every line matched to the PO and goods receipt
OCR reads the header; an e-invoice delivers every line
E-invoicing mandates will push many companies there anyway. In Slovakia, for example, domestic B2B and B2G e-invoicing becomes mandatory on 1 January 2027 (Financial Administration of the Slovak Republic). As a bonus, e-invoices close fraud routes such as a changed bank account on a PDF or a fake invoice sent through another channel.

From practice: what OCR could not fix

In one company, we received about 200 invoices a day for several legal entities, most of them on paper. The mailroom put a barcode sticker on each invoice, and a high-capacity scanner started a new invoice record at every barcode. Efficient at the time, but not automation by today's standards.

Later, in another company, we received about 100 PDF invoices a day. We moved each email to an OCR mailbox, and the system extracted 10–15 header fields. The accountant saw the PDF on the left and the fields on the right, each marked on the PDF, and corrected what was misread. Totals were often taken from the wrong line. When the accountant moved the marker, the system learned a template for that vendor. It improved, but never reached 100%: layouts changed, the number of lines varied and totals moved to another page.

The bigger problem was matching. The system captured the PO number but no lines, so it assigned the whole PO to the invoice. With a partial delivery, the accountant had to undo the match and select the invoiced lines by hand. Invoices with several POs, each partly invoiced, were a nightmare, because the totals never agreed.

Most invoices matched one PO perfectly. But the rework on the rest was so large that the accountants preferred the old way. We set up automated matching and then switched it off.
Without line items, automation stops at the header.
Peter Ilas, FCCA
Senior Finance Manager and Group Director of Accounting and Taxes
With e-invoices, the system receives every line and matches quantity and unit price to the PO and goods receipt, line by line.

Automated approval: a stronger control than a quick click

Many see automated approval as a control risk. But I see it the other way round.

Take a manager who blocks one hour to approve 100 invoices. After closing one invoice and opening the next, that leaves about 30 seconds per invoice, by my estimate: enough to see the vendor and the total, glance at the lines and click Approve if nothing looks suspicious.

An automated approval checks dozens of points on every single invoice: all invoice details, the due date, the VAT/tax rates, the unit price on each line, each line against the PO, and price and quantity against the goods receipt note.

And now my preferred approach: let the system approve 80 invoices and flag 20 for the manager, each with a clear reason: a relatively new supplier, a higher amount than usual or a price difference. The manager can spend about 1.5 minutes on each, check the exact issue and make an informed decision. That is 30 minutes: half the time, with every potential issue reviewed instead of many warning signs missed.
Approving 100 invoicesManager aloneWith automated approval
Invoices the manager reviewsAll 10020 flagged; 80 approved automatically
Time per invoiceAbout 30 seconds (estimate)About 1.5 minutes
Manager's total time60 minutes30 minutes
Vendor and totalSeenChecked
The linesA glanceEvery line
Due dateNo timeChecked
VAT/tax ratesNo timeChecked
Unit price on each lineNo timeChecked
Each line against the PONo timeChecked
Price and quantity against the goods receiptNo timeChecked
What the manager seesThe whole invoice, at a glanceThe exact issue: new supplier, higher than usual amount or price difference
ResultMany warning signs missedEvery potential issue reviewed; an informed decision
It also catches what the accountant missed. Accountants are under pressure to process as many invoices as possible; the system checks every item, every time.

Segregation of duties gets stronger too, because the system is independent of everyone in the company. More approvers are not the answer: in one company, I saw a wrong invoice pass six eyes (team leader, manager and director), because each looked at one aspect only.

I cover the rules and controls that make it safe in detail in the AP Invoice Automation with E-Invoicing course.

Not on day 1: the 4 stages of AP automation

Full automation is not a day-one project, and it never covers all invoices. The optimal approach is through four stages:
StageWhat is automatedWhat people still do
Stage 1Automated intake and validationPeople review and approve every invoice
Stage 2Automated matching and coding, with a human checkAn accountant checks each proposal and improves the rules
Stage 3Straight-through processing for low-risk invoicesFor example, invoices under a set amount that match an approved PO skip people
Stage 4Full automation for all qualifying invoicesExceptions and high-risk invoices go to people; periodic audits test the rules

Why AP automation stalls: compliance only and the old process

I have discussed this with several e-invoicing experts, and they all confirmed what I see: nearly all companies focus on compliance only. They can send and receive e-invoices in the required format, and there it stops. But you can use compliance as the starting point for something that adds real value to your company.

The second reason is the old process. Automating a step needs a new workflow, not the old way with a new tool, which brings no results. When the system prepares the postings, the accountant should review the system's proposals and add what the system does not know: the background. The system follows the rules exactly; the accountant understands them. That is where the accountant adds value.

The business case: cost per invoice and 5 KPIs

Take an AP team of 7 accountants processing 28,000 invoices a year, at a fully loaded cost of $75,000 per accountant:
  • Manual: 28,000 ÷ 7 = 4,000 invoices per accountant, and $75,000 ÷ 4,000 = $18.75 per invoice.
  • Automated: the same volume with 2.5 FTE, or 11,200 invoices each, and $75,000 ÷ 11,200 = $6.70 per invoice.
AP team, one yearManualAutomated
Invoices a year28,00028,000
AP staff (FTE)72.5
Invoices per FTE4,00011,200
Fully loaded cost per FTE$75,000$75,000
Cost per invoice$18.75$6.70
Saving per invoice$12 (−64%)
That saves about $12 per invoice, or 64%. Industry benchmarks show a similar picture: the average AP team spends $9.84 per invoice, and best-in-class teams, who process 1.8 times as many invoices straight through, spend 79% less than their peers (Ardent Partners).

To prove the case, track five KPIs: cost per invoice, cycle time from receipt to posting, first-pass match rate, exception rate and the share of touchless invoices.

Key takeaways

  • Every step from invoice receipt to payment can be automated, but exceptions still need people.
  • You can start with PDF invoices and OCR; without line items, though, automation stops at the header.
  • E-invoices deliver every line, which makes line-level matching and automated approval possible.
  • With automated approval, the manager reviews only the flagged invoices: half the time, and every warning sign seen.
  • Compliance is the starting point; the value comes when you change the process and the accountant's role.

FAQ

Do we need e-invoicing before we automate AP? 

No. You can start with OCR or AI capture on PDF invoices and automate intake and validation. To automate matching and approval at line level, you need structured e-invoice data.

Will we need fewer AP staff?

Usually not fewer people, but different work. Most companies move AP staff from data entry to exceptions, vendor management and analysis.

Is automated approval safe?

Yes, when the rules are right and monitored. The system applies the same checks to every invoice and logs every decision; anything unusual still goes to a person.

Automate your AP process step by step

In E-Invoicing Mastery: AP Invoice Full Automation & Benefits, I take you through all six steps, the controls that make automation safe and the 10 KPIs that show whether it works. 4.5 CPE/CPD credits, 3 hours of video and 60 pages of text notes.

For both payables and receivables, see the program AP and AR Invoice Automation & Cash Flows.

Peter Ilas, FCCA

Founder, RSAgile | E-LITE Consulting
30 years in finance, accounting, audit and internal controls.
Builds practical AI tools for finance teams.
LinkedIn
Created with