How Automation Can Help Agencies Ditch Legacy Systems


Bots come in handy when few other capabilities are available, but they may help agencies bridge to emerging technologies like AI, too.

Robotic process automation has become a critical tool for agencies looking to modernize and divest from legacy systems, agency officials said. 

“[RPA] enables us to pay down our technical debt, to divest from legacy systems and provide the new capabilities where it would otherwise be very difficult to affect those changes,” Chase Levinson, who leads the RPA program at the Office of the Assistant Secretary of the Army for Financial Management and Control, said. “And when you pay down your technical debt when you're investing from systems, it can lead into the strategy that enables modernization.”

Levinson, speaking during a Wednesday event hosted by Nextgov, said his team used automation to cut a process that usually takes five minutes down to 20 seconds. The Army is supposed to move out of the legacy Standard Operations Maintenance Research and Development System, or SOMARDS, by the end of this fiscal year. Without RPA, Levinson said, there was no capability available to make the system change. 

But because there was a small group of workers conducting system changes manually, Levinson’s team was able to automate the process and speed up the work to move away from SOMARDS. 

“It's very hard to find people who can still develop in COBOL, and only a few of them are on staff, it's very hard to do anything other than very basic maintenance for some of these legacy systems,” Levinson said. “So trying to affect some system changes to enable our legacy divestiture was cost-prohibitive. However, we do know that these processes were being done manually by the existing workforce. Well, if we can do them manually, if we have a set business process, we can use RPA.” 

Frank Wood, the supervisory IT coordinator for the Defense Logistics Agency, who also spoke on the Nextgov panel, added that RPA can also help start a migration toward more consolidated architectures. Wood said right now, RPA use cases largely consist of bots bringing up web interfaces, making selections, and entering data. 

But eventually, Wood sees automation technology evolving to be able to access secure application programming interfaces to enable systems to talk to each other through the APIs. 

“So RPA, if you have nothing else, is a very good start,” Wood said. “But as you evolve, you evolve the RPAs and that gets you into the other technologies and the game becomes integration, I think at that point. So I think that's where it's pushing us.”

Some of those other technologies are likely to include artificial intelligence and machine learning. Levinson even said he doesn’t see how to implement AI and ML without using RPA as well.

“I think RPA are the hands that allow more sophisticated techniques, more artificial intelligence sites, tools that affect changes, and actually embed themselves within the business process,” Levinson said. “You don't want to be a brain in a jar, you don't want to have just some kind of prediction engine, sending out a report that no one ever reads. RPA enables that brain to have hands and affect changes and allows us as enterprises to actually achieve savings and efficiencies.”

Getting the workforce comfortable with trying RPA for new and innovative applications takes support from management, though. Shanna Webbers, chief procurement officer at the Internal Revenue Service, another participant on the panel, emphasized the importance of giving teams “top cover” when they’re trying something new and different. 

When Webbers’ team needed to conduct thousands of contract modifications on a short deadline last fall, some were hesitant to use RPA. Previously, Webbers’ office was using just one bot to make data corrections in the Federal Procurement Data System, but it had never been used in another scenario. Letting her team know she had their backs if using RPA for the contract modifications didn’t work out was critical to the success of the effort, Webbers indicated. 

“We need to do intelligent risk-taking, so that we understand what those risks are, and where we're willing to take those risks and where we're not willing to take those risks … And if for some reason it doesn't work out exactly like we thought it was, you know what, then we look for opportunities for lessons learned versus lessons acknowledged,” Webbers said.