Robotic Process Automation: An AR Tale

by Royce Grayson Morse

Let’s say you manage an accounts receivable organization. One of your biggest challenges is cash application; matching payments to invoices is a particularly painstaking task for you and your team. Payments arrive in all sorts of ways: some come with no payment advice, just a company name, which may not exactly match the one in your customer master file. Some come in with unexpected or miscalculated deductions taken, so the amounts don’t match. Some customers get creative and alter the invoice number to include a code only they know the meaning of. Your invoice numbers consist of 12 characters, but several customers have systems that don’t allow that many digits in the invoice field, so you end up with remittances containing incomplete invoice numbers. The list goes on.

You have a hardworking supervisor named Melissa whose job it is to figure all that out and make sure the payments are applied correctly and in a timely fashion. Melissa doesn’t like posting payments “cash on account”; she wants to get them right from the get-go. (It takes too much time to go in after the fact, she says, and sometimes people forget to do it. “Better to do it right than to do it over,” is her mantra.) She’s so dedicated that she often ends up working ten-hour days, along with her team, which is a bad thing. Aside from the overtime cost, Melissa and her staff are overworked and exhausted.

Should you hire more people? Of course, you could. You’ve been doing that for several years now as the business grows. The problem is finding people with the level of precision the job requires and then training them. Every time someone leaves the company—and they do, the job has a high burnout rate—Melissa ends up spending a lot of time hiring and training a new person, time she feels she should be spending doing AR operations and management tasks. You’re getting the feeling Melissa is about over it. You can’t afford to lose her. What to do?

Here’s an Idea…

One option you’ve been considering is RPA, robotic process automation. You know there are several vendors who provide software solutions that use some level of artificial intelligence to do this sort of work. Regardless of how your payments come in, checks, ACH, EDI files, payment advice sent by email attachments, the software can read them, actually figure out what the payments are for and apply them correctly, even if the remittance information is incomplete or a little “off.” That’s the claim, anyway. RPA is supposed to be able to problem-solve like a human being.

It's touted as being a significant step beyond traditional payment automation, because once it’s set up, it’s pretty much hands-off. You know from talking to some of the people in your network that a lot of electronic cash application solutions that are supposed to be automated, really aren’t. Getting the remittances linked to payments and getting them matched to the open AR entry remains a pretty manual process for a lot of them. And that’s not even taking into consideration paper check handling, which requires its own form of coaxing through the remittance process. If what you’re hearing about RPA is true, you think, “we should do it!”

You haven’t mentioned this idea to Melissa, though, because you haven’t yet researched it thoroughly and you don’t want to get her hopes up.

Instead, you mention your idea in confidence to your friend Brad, who works in the IT department. You figure Brad will be able to provide some insight into what’s essentially a software issue. “No way,” Brad says. “As you know, we’re in the middle of implementing a whole new ERP system and this isn’t in the project plan. My boss will never go for that. Even if she does, I just know we’ll break something in the process, and then I’ll be in deep trouble. Plus, we’re really, REALLY busy.”

Later that day, you bump into your boss coming out of a meeting. “Can I have a minute?” you ask him, and he motions you into his office. You tell him about your idea of RPA implementation.

He listens, but he says what you figured he would, “It sounds like it might be a good idea, but I don’t really have money in the budget for something like that. I’ll need to have a really good cost justification to be able to get that past the CFO. Plus, what about IT?”

On your way home that evening, another of your supervisors, Germaine, calls you on your cell and says, “Hey, sorry to bother you, but I heard Melissa has been putting her resume out there. I’m pretty sure she’s getting ready to leave.”

This is bad news, but honestly, you’re not surprised. Melissa has put in some time off requests, and you already suspected she might be lining up interviews.

Game On

That evening, after a little winding down, you boot up your personal computer and start doing your research.

First, the cost of RPA. Your internet research reveals that one software “bot”—a single license—costs about one-third of the annual wages of a person. That’s a fraction of the cost of a full-time AR person! OK, but how accurate will the bot be, and how much work can it get done? You discover that many vendors say they can guarantee productivity increases from 30 percent on up, with a straight-through processing rate of more than 90 percent in many cases. With those numbers, it should be pretty easy to make a business case for the CFO. And cash application is a pretty common use for RPA, so there are several reputable, experienced vendors already out there to choose from.

But what about poor, overworked Brad? How will IT handle the implementation? To your surprise, you discover the IT department probably won’t have to do much, if anything at all.

Implementing an RPA solution is generally handled by operations people, not technical people. It doesn’t require any code to configure it; once installed, the software it learns like a person does, by example. A lot of vendors create their software with business rules built in, which means the bot already knows how to process the most common tasks, like payment application. If it needs additional training, it’s a simple matter of teaching it, often through a drag-and-drop workflow interface: go here, then do that, and if you encounter this situation, do this other thing. It’s a lot like using a common flowcharting program you already use.

It doesn’t interface with Brad’s new ERP system through code; it works just like a person on Melissa’s team does, using the same user-facing tools, the GUI (graphical user interface) and workflows. (Brad will be ecstatic!) Once IT gets the software installed, they can go back to wrestling with their ERP implementation and let you handle the rest.

And Melissa should be thrilled too—after the software is up and running, she should only have to tweak the business rules from time to time, and the bot will work continuously, 24/7, doing those routine tasks you’re now paying people to do, often at time-and-a-half their normal hourly rate. And it won’t make mistakes or get distracted because its significant other is texting it that it has a flat tire.

Implementation

Within a week, you’ve made the case with the CFO, and IT is out of the picture, working away on their big project. Everybody up the chain is on board. While you’re working out the details of selecting a vendor, it’s time to share the plan with Melissa. She’ll be so happy!

Or not. Melissa’s first reaction is confusion. She doesn’t get the “robot” part. You explain it’s not an actual robot, like C3PO, it’s a software tool that thinks for itself through artificial intelligence, learning as it goes. You go on and on to Melissa about how wonderful life will be without having to constantly hire and train new people to do the same routine task, how RPA will work constantly for less money and with fewer mistakes than staff. But you can see she’s still not convinced.

Finally, you stop singing the praises of RPA and actually look at her. “What’s wrong?” Because something clearly is.

She swallows, then answers. “Will it…replace people?”

Aha. You say, “That’s not the intention. The idea is to keep from having to hire new people as the business grows, and reduce the extra hours of our existing staff. I know that’s stressing you out. This way, you can reassign your people to other tasks that fall behind. If the system is good enough, we may not have to replace people that leave. And at some point, yes, we may not need as many clerical people, but business is evolving, right?”

Still, she seems uncertain. It’s time to get real.

“Melissa,” you say, “I know you’re unhappy and you’re thinking about a new job. And if that’s what you need to do, you know I’ll understand. But think about it. You can go somewhere else and do the same thing you’ve been doing here, with the same frustrations and challenges, or you can stay here and be part of the wave of the future. You’ll become our resident expert on RPA. And if other automatable tasks in AR—or even in other departments—come along, people are going to look to you to help them figure out how to use the technology. Won’t that be a career changer?”

Fast Forward

You selected a vendor and the roll out went smoothly. Melissa got on board with the project, helped with the vendor analysis process, and even got a few of her team members working with her to get the business rules tweaked. It took them just a few days to train on the software, and once the program was activated, it was set to run slowly so that they could double-check its work. When Melissa and her crew were satisfied it was performing cash application as it should, it was cranked up to run at full speed. It turned out that after their initial skepticism, the team was all pretty excited about the new technology.

Cash was getting applied faster and with greater accuracy than ever before. When a big customer changed their own system and revised their remittance format, it was not a problem for the software to “learn” to read the new layout. When a brand-new customer came on board and wanted you to download payment information from their web portal, it was easy mode; you and Melissa had that working without breaking a sweat. The little problems that used to drive everybody crazy trying to figure out—the small customer that one day decided not to use its official name on its payments but some unexpected abbreviation, the long-established customer who insisted on using a PO number on its remittances instead of an invoice number, even the customer-specific discount system the sales team put in place (surprise!)—no big deal.

And top management was loving it. Cash was being applied the same day the funds were deposited. Fewer mistakes were being made, and you cut out the overtime. HR was loving it, because the intense recruitment process for Melissa’s team dropped to a fraction of what it had been. And Melissa seemed to be backing off from her energy-drink habit.

Epilogue

The team’s stress level—and overtime—went way down. Your boss was happy, the CFO was happy, and Melissa stopped job-hunting, according to Germaine. She even took the initiative to see how the software could help Jorge, the credit supervisor, make faster and better credit decisions.

And then one day, you get a call from Brad, your friend in IT. “Do you think you and Melissa could come over to accounts payable for a few minutes?” he says. “They’re wondering if maybe you can help them figure out a better way to do stuff.”

Such is the promise of RPA. Of course you’ve heard technology promises before. For a success like that described above, there are prerequisites – see the article Robotic Process Automation in AR—How Close to Reality? Also see Is There a “Robot” in Your AP or AP Department’s Future? New Technology to Keep an Eye On.

Don’t miss the on-demand webinar:

Supervised Machine-Learning in Cash App Automation