What is AP automation and how does it work?

AP automation is software that runs the accounts payable process with little manual keying. It captures supplier invoices as they arrive, reads the data off each one, checks that data against your purchase orders and your ERP, routes anything that needs a decision to the right approver, and posts the finished record into the accounting system you already run. Your finance team reviews the small number of cases that genuinely need a person, and the software handles the rest.

What AP automation is

Accounts payable is the work of receiving supplier invoices, confirming they are correct, getting them approved, and paying them. Done by hand, it means someone opening emails, reading PDFs, keying figures into the finance system, and walking approvals round the business. AP automation covers the software that does the repetitive parts of that work and hands your team the cases the software cannot settle.

The term covers a set of connected jobs that work as one: capturing documents, extracting their data, matching them to purchase orders and goods receipts, running approvals, and posting to the ledger. A tool that handles only one of these leaves most of your AP still manual. The value comes from joining the steps so an invoice can travel from inbox to posted record without being retyped along the way.

It is worth being plain about where the intelligence sits. AI reads the invoice and pulls out the supplier, the net and tax amounts, the line items, and the PO reference. Your team sets the rules and handles the cases the rules cannot decide. The software is the tool. The finance team still owns the judgement.

It helps to place AP automation next to the things it sits beside. Your ERP has a built-in AP ledger that records invoices once someone has keyed them, and automation is the layer that gets the data in and checks it first. Payment automation moves money once an invoice is approved, and capture, extraction and matching all happen before that point. Optical character recognition reads text off a page, and AP automation adds the understanding of what each field means and whether it is correct. The category is the whole run from inbox to posted record.

One thing separates AP automation that lasts from a pilot that stalls: it has to post into the system you already run, with your codes and structure intact. A tool that exports a spreadsheet for someone to import by hand has only shifted the keying to a later step. Posting straight into Sage 200, Sage X3, NetSuite, Dynamics 365 F&O or Infor M3, with the nominal codes and tax groups your finance team expects, is what makes the automation real.

The manual process it replaces

To see what automation changes, follow a single invoice through a manual AP function.

It arrives by email, on paper, or through a supplier portal. Someone prints it or saves it to a shared drive. They read the supplier name, the invoice number, the date, the net, the VAT, and the total, then key those figures into Sage 200, Business Central, or whichever ERP the business runs. If there is a purchase order, they find it and check the numbers line up. If there is not, they work out who raised the spend and why.

Then the invoice needs approval. That usually means an email to a budget holder, a wait, a chase, and sometimes a second chase. When approval comes back, someone posts the invoice to the correct nominal code and department, files the document, and moves to the next one. Multiply that by a few thousand invoices a month and you have a team whose day is data entry and follow-up.

Invoices also arrive in dozens of formats. PDFs, scanned images, email attachments, and EDI feeds all carry different layouts, and suppliers rework their templates without warning. Handling that variety by hand is slow, and it is where mistakes creep in: a transposed figure, a missed credit note, a duplicate paid twice.

The cost of all this is more than the hours at the keyboard. Approvals that stall push work into the last days of the month, so month-end becomes a scramble. Late payments strain supplier relationships and can forfeit early-payment discounts. And when one experienced person holds the knowledge of how invoices really get coded, a holiday or a resignation becomes a risk to the whole function.

For a UK mid-market finance team, the strain shows most across entities and at the close. A group running several trading companies or sites keys the same kinds of invoice into the same ERP under different company codes, and someone has to keep the coding straight for each one. When month-end arrives, the invoices still sitting in inboxes turn into accruals someone has to estimate, and the numbers the board sees are softer than they should be. The more manual the process, the later in the month that picture comes into focus.

How AP automation works, step by step

Under the surface, a well-built AP automation flow moves an invoice through five stages.

1. Capture. The invoice comes in through a dedicated email address, a portal upload, or a scanner, and the software collects it automatically. No one saves files to a drive or sorts them into folders by hand.

2. Extraction. AI reads the document and pulls out the fields that matter: supplier, invoice number, dates, net and tax amounts, and each line. This is where document intelligence earns its place, because it copes with varied layouts without needing a fixed template for every supplier.

3. Matching. The software compares the invoice to its purchase order and, where goods are involved, the goods receipt. A clean match can post without anyone looking at it. A mismatch, such as a price or quantity that does not agree, is flagged for review.

4. Approval. Invoices that need a sign-off route to the right person against your rules: value thresholds, cost centres, and delegation when someone is away. Approvers see the invoice and its context in one place, so a decision takes seconds.

5. Posting to the ERP. The finished record posts into your accounting system with the nominal codes, departments, and tax groups intact, and the original document stays attached for the audit trail. Stratas connects to the ERP you already run, across 100+ ERP integrations including Sage 200, Sage X3, Business Central, NetSuite, OrderWise and more.

Two paths run through those stages. The straight-through path is for invoices that match cleanly and fall within the rules, so they capture, extract, match and post with no one touching them. The exception path is for everything else: a missing PO, a price variance outside tolerance, a supplier the system has not seen before. A good setup sends as much as possible down the straight-through path while making the exception path quick to work.

AI and rules do different jobs across those stages. AI handles the reading, coping with layouts, wording and the ways real suppliers format a document. Rules handle the deciding: which invoices can post, which need a second pair of eyes, and who signs off at what value. Keeping those responsibilities clear is what stops automation from becoming a black box your finance team cannot trust.

Matching is the step that does the most to protect your cash, and it is worth understanding on its own. For the detail, read what 3-way matching is and how it catches overpayments before they leave the building.

What AP automation does not do

Honest expectations matter here, because oversold automation is how implementations disappoint. A few limits are worth stating plainly.

It does not reach 100% accuracy. Stratas publishes 89% field-level extraction accuracy, measured on real customer documents: on day one, roughly nine fields in ten come through correct. Validation then checks the extracted data against your records, and anything that fails a check goes to a person, which is why exception handling is built into the design from the start.

It does not remove your finance team. The team spends its time on the invoices that do not fit, on managing suppliers, and on closing the month. The people who understand your business still make the calls the software cannot.

It does not fix bad master data. If your supplier records, PO discipline, or nominal structure are messy, automation will surface that quickly. Clean data upstream is what lets the software post with confidence, and part of a good rollout is putting that right.

It does not make your buying decisions. AP automation checks that an invoice matches an order that already exists, and the purchasing controls that decide what gets ordered, and by whom, sit upstream in procurement. Automation makes weak controls there more visible, and it rewards teams that have their PO process in order.

It does not run itself from day one. The rules that decide what posts automatically and what routes for review are set up around your process, and they sharpen as the software learns your suppliers.

The benefits, and the numbers behind them

The case for AP automation rests on cost, speed, and control. The numbers below come from independent research.

On cost, widely quoted industry estimates, usually attributed to the research firm Gartner, put the price of processing a single invoice by hand at between £4 and £25, and higher again when an invoice goes wrong. The original studies are hard to trace, so treat that as a rough range; our guide to what it costs to process an invoice shows how to work out your own figure. Most of that cost is labour: the reading, keying, chasing, and fixing. Automating the routine invoices takes the bulk of that work out.

On speed, Ardent Partners' AP Metrics that Matter research found that the fastest AP teams process an invoice in 3.1 days on average, against 17.4 days for everyone else, and they do it at roughly a fifth of the cost per invoice.

Those faster cycles change what is possible with supplier terms. Many suppliers offer a small discount for settling early, and a manual AP function rarely captures them because invoices are still working through approval when the discount window closes. When an invoice is captured, matched and approved in days, the team can take those discounts as a matter of course, and the saving builds up across every supplier that offers one.

Cost also falls because there is less rework. A figure keyed wrong has to be found, queried with the supplier, corrected and re-posted, and each of those steps costs time that never shows on the original invoice. Reading the data off the document at 89% field-level accuracy, then validating it against the ERP before it posts, takes most of that rework out of the month.

On control, every invoice carries its own audit trail: who approved it, when, and against which PO. That trail is also a defence against the ways money leaks out of AP. Duplicate invoices, inflated quantities, and invented suppliers are the common routes, and they are hard to spot by eye across thousands of documents a month.

Automated matching checks every invoice against a real purchase order and, where goods are involved, a goods receipt. Duplicate detection flags anything that looks like a repeat before it is paid, and the record shows who touched each invoice at every step. For a finance director, that means the controls can be evidenced at audit, with a record for every invoice, and fewer surprises at month-end.

There is a benefit that does not show up in a cost-per-invoice figure. Nobody joined a finance team to key invoice numbers or chase a manager for a signature. When the routine work is handled, the same people can spend their time on the things that need judgement: querying odd charges, managing cash, and giving the business real numbers sooner. The headcount you have goes further as volumes grow, without the AP team having to grow at the same rate.

How to know you need it

Not every finance team needs to automate today. These signals suggest the manual process is costing more than it should.

• Your AP team spends most of its time keying invoice data and chasing approvals by email, so skilled people are doing work that does not need their skill.

• Invoice volumes are rising, and the only way you can keep pace is by adding people or letting a backlog build.

• Exceptions, duplicates, and mismatches are common, and each one pulls someone off their other work to investigate.

• Month-end is slow because invoices are still sitting in inboxes waiting to be posted, so the close leans on estimates.

• You have missed early-payment discounts, or paid a supplier late, because an approval stalled somewhere in the chain.

• An audit meant hunting for documents across shared drives, email, and filing cabinets, with no single record of who approved what.

None of these on its own means you must act today. Taken together, they describe a process that is absorbing more of the working week than the work is worth, and one that grows more fragile as the business grows. The cost of leaving it is quiet, which is why it rarely forces a decision on its own.

If several of these are familiar, the question is which part of the process to fix first. The answer usually starts at capture and matching, because clean data at the front makes every later step easier.

A sensible first step is small. Take a representative sample of your invoices, the awkward ones included, and see how much of the data the software reads correctly and how cleanly it matches to your orders. That tells you more about the fit for your business than any feature list, because it runs on your own documents and your own ERP. That is the test that predicts how the software behaves once it is live.