Robotic process automation  (Alexander Supertramp/Shutterstock.com)

When not to automate

Participants

Jimmy Chen
Manager, Department of Veterans Affairs

Laurie Cook
Vice President, Public Sector Sales and Alliances, Blue Prism

Marina Fox
DotGov Domain Services Program Manager, General Services Administration

Sanjay Gupta
CTO, Small Business Administration

Keith Nakasone
Deputy Assistant Commissioner, Acquisition, Office of Information Technology Category, Federal Acquisition Service, General Services Administration

Theodore Woronka
Assistant Director, Finance & Administration, Department of the Interior

Ranjeev Mittu
Branch Head, Naval Research Lab

Michael Fairless
Branch Chief, Securities and Exchange Commission

Mark Krzysko
Director, Acquisition Data DASD (AE)/Acquisition Analytics and Policy, Department of Defense

Margaret Moon
Financial Analyst, National Science Foundation

John Powers
Associate Director, Information Security Oversight Office
National Archives and Records Administration

Lt. Gen. (Ret.) Harry Raduege
Senior Counselor, The Cohen Group

Howard Spira
CIO, Export-Import Bank of the United States

A group of government experts recently gathered to discuss the challenges involved in adopting automation, including convoluted business processes and an incomplete understanding of how best to use the technology. Participants agreed that automation is essential to a customer-centric, digital government, but not every manual or error-prone process can be fixed with robotic process automation or artificial intelligence.

The discussion, facilitated by GCN's sibling site FCW was on the record but not for individual attribution, and the quotes have been edited for length and clarity. Here’s a sample of what the group had to say. The full version is available on FCW.

There was a lively discussion about taking advantage of the opportunity to evaluate and improve processes before deciding whether to automate them.

“Several years ago when the executive order on reducing the size of the federal government came out, I realized there was no way we were going to be able to even come close to meeting that financial target,” one executive said. “So I convened my teams and said, ‘We don’t have the option of saying we can’t deliver the service so let’s come up with the attributes for an idealized system. What would it look like?’ And one of the things that we discovered is that focusing on individual administrative processes is a bad mistake because it is a single function. And it’s one of many things that everyone does.”

To solve the problem, the participant added, “we began by thinking of our employees as customers. There are 2 million of them around government, and I think they are really the forgotten bunch. We tried to look at it from their perspective.”

Others also said that finding the right balance between people, process and technology is a key element of any automation effort. One executive described an effort to automate a business process that was spreadsheet-based. “We started out thinking it was pretty straightforward. But then we looked at the data. It was coming from different systems and it was conflicting, it was in different formats, pieces were missing. We had to back down and say, ‘We’ll automate the things that work well, and the user is going to be integrated systematically to fix things that automation will never be able to solve because the data is so poor.’ That was the way we got acceptance.”

“Once we’ve identified an idea, we start small, do a proof of concept, demonstrate the value, and see the feasibility of the technology and the business processes,” another executive said. “If you just see technology, a lot of people will go after the shiny object in front of them. But a critical first step is looking at those processes and figuring out what shouldn’t be automated because not everything should be.”

“In my agency, we automated and digitized processes, but we never took the time to make sure those steps were no longer needed,” a different participant said. “We have tried to address it by having a dialogue about the process side of it and asking: ‘What are you trying to do with this? What do you get out of it? And is there another way you can do this?’”

The need to shift the conversation was a common theme. “Everybody says, ‘I want machine learning and AI,’” one noted. “But they don’t necessarily need that, and even if their problem is oriented to that solution, do they have the subject-matter expertise to take advantage of it? I spend 90% of my day governing the data because I want to give them authoritative, real data. Many people don’t understand that. They think just by putting AI on a bunch of data, it will clean itself up. But it won’t. We’ll just have bad AI projects.”

Note: FCW Editor-in-Chief Troy K. Schneider led this roundtable discussion. The Sept. 11 gathering was underwritten by Blue Prism, but both the substance of the discussion and the recap on these pages are strictly editorial products. Neither Blue Prism nor any of the roundtable participants had input beyond their Sept. 11 comments.

Read the full discussion here.

About the Author

Terri Huck is FCW's print managing editor.

Stay Connected

Sign up for our newsletter.

I agree to this site's Privacy Policy.