How do you automate invoice processing, and what about the exceptions?

Supplier invoices arrive in a shared inbox as PDFs, get opened by someone in finance, typed into the accounting system, matched against a purchase order by eye, and chased for approval. On a quiet week it's a few hours. In the week before month end it's the whole job, and the mistakes made at speed come back as credit notes and awkward calls with suppliers.

This is the process most finance teams hand over first, and it's a good choice: invoices arrive every day, they follow the same shape, and the decisions that need a person are easy to describe. What follows is how the automated version works, where it still needs you, and how to set it up so it doesn't create a new kind of mess.

What the system does with each invoice

Think of it as four steps that run on their own, with a person at the end of two of them.

1. Capture. The invoice arrives by email, through a portal or as a scan. The system picks it up, records where it came from and when, and reads the fields: supplier, invoice number, date, line items, amounts, tax, payment terms.
2. Match. It finds the purchase order and, where you use them, the goods receipt. It compares the fields you've said matter and marks each one as agreeing or not.
3. Route. If everything agrees within the tolerances you've set, the invoice goes into the approval queue for the right person, or is approved on a rule if you've allowed that for small, routine amounts. If something doesn't agree, it goes to the person who owns that kind of exception.
4. Post. Once approved, the invoice is posted to accounting with its coding, the purchase order is updated, and the file is stored where your auditors expect to find it.

Reading the invoice is the part people worry about and the part that causes the least trouble. Modern document reading copes with different layouts, scans and the occasional photograph of a receipt. The step that decides whether this works for you is the second one, and it depends on rules only you can write.

Decide what counts as a match before anything is built

Sit down with the person who does this today and write out what they check. It is usually a shorter list than they expect, with a few judgement calls hiding in it.

• Which fields must agree exactly: supplier, purchase order number, currency.
• Which fields can differ within a tolerance: a small rounding difference on the total, a quantity that's within what was received.
• Which differences are normal and shouldn't block anything: a freight line that isn't on the order, a price that moved within an agreed range.
• What happens when there is no purchase order at all. Some suppliers never get one. Decide whether those invoices go to a person every time or follow a simpler rule for a known list.

Write these down as plain sentences with the actual numbers in them. They become the rules the system runs on, the test cases it's checked against, and the document your auditor will ask for. If you can't agree a tolerance, pick the strict one to start with. Loosening a rule later is easy; tightening one after a few bad payments is not.

What to do with the invoices that don't match

Every invoice process has exceptions, and the automated one surfaces them faster than the manual one did. That feels like more work in the first week, because you're now seeing all of them at once instead of finding them months later. It settles quickly once two things are in place.

The first is that each exception has an owner. Price differences go to the buyer who placed the order. Quantity differences go to whoever receives goods. Unknown suppliers go to accounts payable. Nothing lands in a shared queue that everyone assumes someone else is watching.

The second is that the exception arrives ready to decide. The person opens one message and sees the invoice, the order, the receipt if there is one, and exactly which line differs and by how much. They approve, reject or send it back to the supplier from there, and the system records what they chose. If they have to go and find the paperwork, the automation has saved nothing.

Over the first months, look at which exceptions keep coming up. A supplier who always adds freight is a rule waiting to be written. A buyer whose orders never match is a conversation. The exception list is the best map you'll get of what's loose in your purchasing.

Where a person stays in the process

Two places, and they should be visible to everyone who touches the process. Exceptions, as above. And approval of anything above the threshold you set, even when it matches perfectly. The threshold is a business decision: some companies let a clean match under a certain amount post on its own, others want eyes on everything for the first quarter. Which approvals can run on rules and which need a person is a decision worth making deliberately rather than inheriting from the old process.

What a person no longer does is type. Nobody re-keys a supplier's address, nobody copies a total from a PDF into a field, nobody hunts through the inbox for the second page of a scan. That is where the hours go, and it's the part that was never a good use of a trained person's day.

How to start without disrupting month end

1. Pick one entity and one kind of invoice to begin with: the suppliers who always send a purchase order and whose invoices usually match. Leave the awkward ones for the second phase.
2. Take a baseline. Count the invoices processed last month, the hours spent on them, and the number that needed a correction after posting. Keep the figures; you'll want them in three months.
3. Run the system alongside the manual process for a few weeks. The same invoices go through both, and you compare. This is where the match rules get corrected.
4. Switch the matched invoices over first. Exceptions can keep going to the same people they always did, now with the documents attached.
5. Widen to the next supplier group once the exception rate for the first is steady and understood.

If you're not sure invoices are the right place to start, what should you automate first sets out four questions to put to any process. Invoices pass all four for most businesses, which is why they come up so often.

What to measure

Agree these before the build and take a reading before anything changes, so the result is a comparison and not an impression.

• Cost per invoice processed: the hours spent on invoices each month, costed, divided by the number of invoices.
• Hours of manual work removed each week, from the people who used to do the typing and the matching.
• Errors and re-work: corrections after posting, duplicate payments caught, credit notes raised because of a processing mistake.
• Time from invoice received to approved, which is what your suppliers notice and what early-payment terms depend on.

These are the measures on our Operational Efficiency page, and they are the ones we report against after launch. The first call is a good place to talk through your invoice volumes, how matching works for you today, and whether this is the right first system or whether something simpler would do.