Putting on your AI oxygen mask: automated GL codes for non-PO invoices

by AppZen November 18, 2022

Accurately coding general ledger (GL) accounts for invoices missing purchase orders (POs) is harder than climbing Mt. Everest while holding your breath. From code category guessing games that screw up budgets to outright embezzlement, the accuracy of your spend data and even the safety of the organization could be compromised if your team gets cost centers and GL codes wrong. Here’s how to fully automate the process in near-real time using artificial intelligence.

It’s okay to breathe, now. You’ve got this.


The view from the top: what fully automated GL coding looks like

The company was overwhelmed–none of their invoices had an associated PO. Without any context, manually researching and applying the correct general ledger codes to these transactions was costing them time and money. Their existing automation tools weren’t helpful and outsourcing was too expensive. They jumped at the chance to let AI do all the work. AppZen’s Autonomous AP seamlessly integrated with their existing ERP to apply artificial intelligence with deep finance knowledge to their accounts payable process.

In just one, 2-week period, 76% of the company's 1100 invoices
were automatically coded to the correct GL account, leaving
only a fraction for their finance team to handle.


Manually matching POs to the general ledger is like a headache at altitude

General ledger account codes are assigned to describe different types of transactions, in order to make accounting data easier to organize and report. POs give your finance team information about who is purchasing items, what for, and from which cost center or business unit so that GL codes can be assigned. In this way, purchases are attributed to the correct budgets. Used together, GL codes and POs help you can track spend across your organization at various levels, from broad to granular categories. Or so you hope.

Without a PO, AP clerks have no idea what the invoice is actually for, whether the company was charged correctly, or whether the goods or services were approved and received. Limited options in procurement and ERP systems force requestors to select a predefined spend category. Guessing the spend category, which drives the GL code, can lead to misclassification. Even using an automation system is no guarantee, as it may also incorrectly code invoices (“office supplies” > should be “computer equipment”). And it can’t learn from its mistakes, meaning you have to step in and make adjustments. While all of this is going on, the invoice goes unpaid. This creates a lot of confusion and prevents your finance team from controlling where the money is going.

Some organizations will outsource the classification of their spend just to be sure they’re not faced with the serious consequences of incorrectly-coded data. And those with a high volume of service-based invoices may require additional time to identify line-item levels, spend categorizations, finance taxonomies, and GL codes. All of this leads to late fees, unhappy suppliers, and stressed-out AP teams with little to no guidance on how to code and process invoices.


Your oxygen mask: artificial intelligence

More and more finance teams today are using the latest artificial intelligence tools to fully automate important data-gathering functions like cost center attribution, GL codes, and spend data classification. They’re reaping the benefits of these systems’ ability to quickly and accurately handle large amounts of information. No more outsourcing, manual reviews, or other staffing challenges when scaling up or shifting gears. And the software easily shares information with ERP systems for a smooth handoff.


The AppZen path: how an AI review of invoices works

Although manual matching can be a very human-intensive part of the AP process, there are ways to make it a bearable climb. Scaling up your invoice automation is easy with AppZen’s Autonomous AP solution. Built using artificial intelligence software with deep finance expertise, it understands your processes and can learn your unique AP environment in just a few short months and support every member of your finance team. So how does it work, exactly?


It all begins when your supplier’s invoice is emailed to your AP inbox. Using tools such as computer vision and semantic understanding, artificial intelligence software reads and pulls out the invoice header and other information from the PDF. It then enters that information into the system.

And here’s where AI is unique: it knows what it’s reading.

It understands the context of the information and can make intelligent predictions and accurate decisions about what it’s looking at. If it’s not confident that its prediction is correct, it asks you to double-check it. And the answer you give helps it to learn–after just a few examples, it no longer has to ask for your help.


Breathe easier: using AI to accurately predict GL codes

Using an AI-first approach right from the point of extraction, Autonomous AP predicts and accurately assigns cost centers and account codes specific to your chart of accounts (COA), based on your historical coding data and data it has extracted from other invoices. Custom data sets can even be checked to confirm these different segments using other information it finds on the invoice, such as an employee ID that maps to a specific location or cost center. And it does it all in near-real time and with 100% confidence.


The AppZen difference

With Autonomous AP, most invoices move from “extraction” to “ok-to-pay” with no human intervention, improving your invoice processing efficiency by up to 80%. It also detects and drives down duplicate spending by up to 5%. 

When an invoice comes into the AP Inbox it's first checked for duplicates across invoices, card transactions, and expenses. It then moves to PO matching and the assignment of GL codes and ensures that all payouts across your systems meet your spend policies. Finally, it enters the procurement or ERP system, all without human intervention. Our intelligent AI software learns quickly and automatically adjusts its predictions based on just two to three data samples of user feedback.

Reach the height of processing volume, save money, and help your company scale that next section of the mountain without the need for additional staff. Take a deep breath and contact AppZen for a demo today to learn how our Autonomous AP solution can help you.


BONUS: Questions to ask vendors

Here are some questions to ask when evaluating the best solution for your company:

  • Can your solution provide GL codes and cost centers for my non-PO invoices to be validated? 
  • If a PO is missing from an invoice that requires a PO, can your solution find the correct PO with a high level of accuracy?
  • If the PO is missing and your system can find it, can it then find the correct GL codes and cost centers without my help?
  • Can your system self-correct its mistakes, using only a few examples, so it codes correctly in the future?
  • Can you show me exactly how your system does all of this?