Defense Logistics Agency has ‘secret sauce’ for deploying software bots
Roughly three years into their journey deploying software-based bots — including many that are unattended — Defense Logistics Agency officials are observing big impacts they never predicted early on.
The agency is leaning more and more on robotic process automation (RPA) to automate repetitive tasks and unburden its workforce.
“I firmly believe that every organization has something that can benefit from RPA,” DLA’s Procurement Process and Systems Division Chief Rusty Wells told FedScoop. “What we kind of did at DLA was like, ‘hey, if something’s not coming to mind, go ask your end users what frustrates them. What do they hate to do as part of their daily job?’ Because I guarantee you, they’ll probably tell you. Then start small, grow the mindset. You may not get it exactly right. You might fail a little bit. You may not be able to automate the whole process. But you’re going to learn from something, you’ve just got to get started.”
In a recent interview, both Wells and DLA’s Robotic Process Automation Program Manager, Frank Wood, briefed FedScoop on what makes their agency’s automation-pushing program so unique, and how and where it’s going.
More than just hours saved
Operating as the nation’s combat logistics support agency, DLA manages the end-to-end global defense supply chain for all five military branches, 11 combatant commands, and other Defense Department components. The organization, which has about 26,000 employees deployed globally, buys about $42 billion in goods and services annually to enable the Pentagon’s missions, including food, clothing, equipment and information technology.
The agency is increasingly turning to RPA to automate repetitive tasks and free up staff to execute on such massive volumes of business.
“Today, we’re up to 136 use cases built and that includes, by the way, 36 enhancements to original use cases to which additional capabilities have been added,” Wood said.
He noted that such enhancements don’t involve “just a tweak,” but actually go through the same governance and software development lifecycle process as any new use case. RPA software firm UiPath supports DLA’s many deployments.
Wells highlighted some of those evolving RPA applications that he’s seen making waves from his perch leading procurement.
“At DLA, we do a lot of contracts. In fiscal year 2021, we issued about 10,000 contract actions per day. So, going over a course of a year that’s roughly 3.7 million actions,” he said. “When you have that many actions, even if you have a small percentage of things you can’t get to, it adds up quickly. So, anything we can do to automate some of those manual steps is definitely beneficial.”
A relatively new bot developed and unleashed for his team helps alleviate some of the annoyances associated with closing out long-term contracts at the end of their duration — tasks that Wells noted are “not exactly the most glamorous part of the job.”
Essentially, the bot they produced identifies scenarios where contracts can be closed out automatically, without the end user or the contract administrator having to take action because all the delivery orders are paid and there are no unliquidated obligations.
“We could have this bot go in and take those off the work that people would have to go in and manually do,” Wells noted, adding that “RPA is definitely helping us along the way here.”
Another bot was created to help ease some of the challenges that surfaced when the COVID-19 pandemic hit. It concentrated on post-award requests (PARs) that DLA handles, where suppliers can submit requests to contract administrators when there are potential concerns or issues after the contract has been awarded.
“When the pandemic started two years ago, we said you know, we probably need to do some tracking of anything related to COVID-19,” Wells said.
The team came up with an idea to create a new and specific code for PARs related to supplier performance of contracts impacted by COVID-19 and then have an automation tool go in every day and look for the certain coded elements in new or in-process items. It would report those details to selected personnel within the agency so that they could have more advanced notice about issues arising. If contracts were critical, they could then see and address them more quickly.
It took about one month to implement that RPA tool, which is still up and running today.
About 160 transactions a week were flowing in with that special code after the launch. Now, there’s additional capabilities that bot performs, like opening the request, pulling the information into a spreadsheet and sending it to the humans that can address it.
“If you think about that automation, it expanded our ability to do something we weren’t touching before because we just didn’t have people doing it. So, it was kind of a new area for us and it’s still running. And when we looked at what we thought it would do for us, I think we’ve gotten about 10 times the benefit that we originally thought we were going to get from it,” Wells said.
While reporting on RPA use cases often centers on time saved by the technology, Wood noted that “there’s more to it than just the number of hours saved.” He said there are instances where automation is enabling a workflow to survive a technology-driven increase, like an uptick in transactions.
“Also, it could be an enabler to do a workflow that’s not otherwise possible, for which there is no manual workflow today. I’m doing a couple of use cases where they need something done — and if we don’t do it, it’s not going to get done, and it’s an actual programmatic workflow,” he said.
He continued: “And then there’s integration, right? You have two systems that need to be integrated. They are integrated today by a human who does a swivel chair between two web interfaces, say. And of course what we’re doing there is we’re integrating those two systems, making sure there’s communication and information. That is either not possible without RPA or even if it was possible some other way, that other way is, you know, untenable in terms of expense or the time it would cost to get there,” he explained.
Wood added that it’s important to remember that RPA lives in a continuum of automation technologies.
“The RPA that does a job today may not be needed next week for all the right reasons. We’re very sensitive to that,” he said.
Of the agency’s 136 bots, 123 are running in an unattended manner.
“That means it’s being executed automatically by a software entity — a digital worker — not a human,” Wood said.
The attended bots are run by a human, via a laptop, using credentials that come from their common access card (CAC). DLA’s unattended bot servers have their own credentials within the agency’s active directory, according to Wood.
“About 96% of our population of use cases are unattended,” he said. “That’s highly unusual within DOD and all of the government. My fellow programs out there, at least in DOD, I think there’s a total of about four or five unattended use cases out there. I have 123 within my program.”
So many unattended bots require the agency to implement much tighter IT controls in general, and cybersecurity controls in particular.
There are multiple approaches to using automation, and Wood emphasized that “one size or style doesn’t fit” every organization. But in speaking to colleagues at other federal agencies, he’s found that “the reason we’re so far ahead of other folks is we actually start from the center” by establishing — from the beginning — an enterprise RPA effort built around compliance upfront.
“What do I mean by that? Our particular platform instance is what we call on-prem. It’s within our own virtual enclave. So, our virtual machines which comprise our platform run within a DLA enclave,” Wood explained.
The team is planning to migrate that over to the Defense Information Systems Agency eventually, but for now they control it in a logical and physical network environment. Logical network environments are basically virtual representations of all or some of the physical network environments.
DLA also implemented a hardware security module solution that interacts with the bot servers within the enclave to give them the same software certificate level of credential that a human would have with their CAC.
“So that’s the ‘secret sauce’ — to do the network stuff, physical and logical, and have the identity stuff solved through a hardware security module. But you’re not done there,” Wood added.
There are also processes in place to ensure that the deployed bots have the identical restrictions that humans would have in accordance with controls to complete certain transactions in many, often financial, systems.
All of this automation is unfolding as DLA is going through a much broader digital business transformation.
Within that context, Wood noted that officials also formed a DLA steering committee that represents and involves all of the agency’s constituent organizations, as well as process owners. Once an automation goes through the software development lifecycle and reaches technical feasibility it goes into a backlog that is prioritized and voted on by that committee. From there, other steps are taken, including running the bot through a series of testing code standards. Then it’s put into production by a configuration management working group.
“So, that bot doesn’t do anything that everyone hasn’t agreed it should do. That further assures the bot. If you think bots are out there, you know, like Hollywood [movies show] doing what they decide — they can decide nothing. They haven’t the intelligence. What we do tell them to do is under great oversight through the central program, above and beyond any of the technical stuff we talked about,” Wood said.
Wells is a member of the steering group.
“We get bot ideas from top to bottom within the organization. Some of our senior leaders have commented on things and it gets people thinking about where we can automate,” he noted.
Looking to the future, Wells is eager to find new opportunities involving RPA that benefit not only DLA’s personnel, but also the agency’s broader supplier community and close industry partners that run similar operations.
Wood is excited to continue building momentum and to grow the existing RPA platform in volume — and intelligence.
“Right now, we do RPA, which is a very simple technology. It actually mimics, in a robotic fashion, processes that a human does. It is not very smart. But that doesn’t mean you can’t make it smart. So, our antenna is up all the time to integrate with other technology areas within DLA, within that digital business transformation context. As they develop additional capabilities, we’re looking for ways to kind of drive more cognition into our platform. So, our ears are on and we’re excited that that day is coming very, very soon,” Wood said.