<img src="https://ws.zoominfo.com/pixel/UN1mC4S8pYDBTzsJVYoE" width="1" height="1" style="display: none;">

What goes wrong in AP automation projects

When an AP automation project stalls, it usually gets stuck on decisions the company wasn't ready to make. Exception rules, approval paths and the different ways invoices arrive all need answers, and the people who can answer them often weren't involved early enough. The projects that go well bring every stakeholder in at the start, accept that some exceptions will be handled by hand at launch, and leave real time for testing before go-live.

Most companies that automated AP are still keying invoices

AP automation has been around a long time, and manual work still hasn't gone away. IFOL's Accounts Payable Automation Trends 2026 report, summarized by SAP Concur in June 2026, found that 77% of organizations still enter invoices into their accounting systems by hand, up from 66% a year earlier. Only 7% describe their AP function as fully automated.

Ardent Partners' State of ePayables 2025, sponsored by Bottomline, points to where the time goes. Delayed approvals (49%) and high invoice exceptions (48%) remain the top barriers to efficient processing. Buying software doesn't settle either one. Someone has to decide how approvals and exceptions should work, and then the system gets built around those decisions.

That lines up with what we see in our own projects.

Trying to automate every exception

The most common reason one of our projects drags on is a customer who wants the system to handle every exception perfectly on day one.

It's an understandable instinct. Once people see what the workflow can do, they start listing every odd invoice they've dealt with in the last five years and want a rule for each one. Sometimes they start designing new exception processes that never existed before. Every new rule has to be designed, built and tested, and the go-live date keeps moving.

A better approach is to let the workflow handle the bulk of your invoices, roughly 80% as a rule of thumb, and route the rest to an exception queue where your team handles them by hand. Once you're live, you'll see which exceptions actually come up often enough to be worth automating. Trying to make everything perfect at launch is one of the surest ways to stall a project.

Getting everyone who touches an invoice involved early

Before kickoff, we ask customers to identify every key stakeholder and bring them into the early planning, even people who aren't part of the accounting process. Not everyone needs to sit in every meeting. Everyone should understand, at a high level, how an invoice moves from start to finish, and that includes the approvers out in the field.

We recently talked with a company that learned this the hard way. Accounting chose an AP solution on its own and tried to implement it on a tight budget. The result felt disjointed and incomplete. The approvers were project managers in the field, the same people who bring in the company's revenue, and they didn't like how the system worked. The two groups never got on the same page about what the system needed to do, and the company ended up looking to start over.

Testing takes longer than most teams plan for

The other thing customers most often underestimate is the time needed for thorough use case testing as the project nears go-live.

Testing is where real-world scenarios meet the new process, and it's usually where we find the exceptions nobody thought of during planning. A vendor that always bills a certain way. An invoice that needs a second approver. A key supplier that sends invoices through its own portal instead of by email. If those cases aren't tested, they show up in production instead, where they cause delays and frustrate people who were told the new process would make their day easier.

Plan for testing to take real time from the people who know your invoices best, and keep it out of month-end weeks if you can.

What the first month after go-live should look like

When the work up front was done well, the first month brings a handful of small enhancements and adjustments. Nothing that requires major rework. That's normal and expected.

When something bigger does surface, it's usually an exception that never got tested, often tied to a specific vendor, or a way invoices enter the process that we were never told about. Those discoveries don't necessarily mean a project is in trouble. They do make everyone step back and rethink part of the process at a point when the team expected to be done.

How much should you trust your touchless rate?

Most AP systems are measured on their touchless or straight-through rate: the share of invoices that post without anyone handling them. It's a reasonable goal, and the time it saves is real.

A high touchless rate is only worth as much as your confidence in it, though. Before invoices flow straight through, you need guardrails, like matching rules and duplicate checks, that let your team trust what posts without them. If your AP manager can't explain why a particular invoice went through untouched, the number deserves a closer look before anyone celebrates it.

Where AI fits in AP automation now

AI is starting to change the AP automation landscape, and the biggest shift so far is in capture. For years, capture platforms relied on OCR and machine learning models to read invoices. Those are now being replaced by AI-based capture engines, which is a very different process. We're also starting to see AI build new workflows and processes from nothing more than a plain-language conversation. The next few years will be interesting for AP automation.

That doesn't mean everything you've already built needs to be reworked around AI right away. Legacy routing and approval tools in particular work very well. If you want to start somewhere with AI, look at the newer AI capture tools first, especially if you've implemented automation and still find your team keying a lot of invoices.

How long should a project take?

Most of our AP automation projects take four to six months, and some have been done in as little as eight weeks. The pace is usually set by how much time your team can give to decisions, testing and training while still paying vendors and closing the month.

What to do this week

If you're planning an AP automation project, or restarting one, a few things are worth doing before you talk to any vendor, including us:

  • List every way invoices arrive today: mail, email, supplier portals, and invoices handed to someone in the field.
  • Name everyone who touches an invoice from receipt to payment, including field approvers, and plan to bring them in early.
  • Pull last month's exceptions and sort them into the ones that come up often and the ones that are rare. The rare ones can stay manual at launch.
  • Decide who will test, and block time for them in the weeks before go-live.
  • Collect a set of real invoices, including the difficult ones, to test with.

Frequently asked questions

Why do AP automation projects fail?

Usually because decisions about approvals, exceptions and invoice sources weren't settled before the build, or because the people who use the system weren't involved in designing it. Incomplete testing before go-live is the other common cause.

Why are we still entering invoices by hand after automating?

Often the system was set up for the most common invoices, and anything outside those rules falls back to manual work. IFOL's 2026 research found that 77% of organizations still key invoices by hand. Reviewing which exceptions come up most often is a good place to start.

How long does AP automation take to implement?

Most of our projects take four to six months, and some finish in eight weeks. Your team's availability for decisions, testing and training usually sets the pace.

Should we automate every exception before go-live?

No. Let the workflow handle the bulk of your invoices and route the rest to an exception queue your team works by hand. Automate more once you see which exceptions actually recur.

Should we rebuild our AP automation around AI?

Not right away. Routing and approval tools that already work well can stay as they are. If your team is still keying a lot of invoices after automating, the newer AI capture tools are the best place to start.

Who needs to be involved?

Everyone who touches an invoice from receipt to payment, including approvers in the field, at least at the level of understanding the process start to finish. Not everyone needs to be in every meeting.

Let's map out how your invoices move today

If you're weighing an AP automation project, start with how invoices and approvals move through your process now. Our AP Assessment points you to the areas to look at first, or talk with us and we'll walk through it together.