Breaking Down Artificial Intelligence Barriers

Building upon a successful Proof of Concept testing the application of Robotic Process Automation technologies, the U.S. Army Joint Munitions Command cleared a major hurdle to wide-scale deployment of autonomous RPA digital workers.

JMC recently secured approval to field fully automated digital workers using non-person entity credentials on Army networks. The new approval applies to any Army network within the 7th Signal Command’s Area of Responsibility and allows for theater-wide reciprocity for Army use within the Continental United States.

According to Tony Crossen, Chief of Integration and Compliance within the JMC G2/6, obtaining the approval was essential. “In order to achieve the most effective implementation of RPA, the digital workers would need to be able to be unattended – meaning they would be assigned their own credentials to access the Army networks and systems required to perform their assigned tasks,” said Crossen. He further explained that “this is important because it allows the digital worker to perform tasks without the credentials and equipment resourced to a human worker, meaning the human worker is free to perform other duties.”

Connection Between People And Artificial Intelligence Technology, Human And Robotic Hands (Photo Credit: Courtney Maxson, Joint Munitions Command)

Once approved, JMC wasted little time in deploying digital workers. JMC previously identified 46 possible AI use cases, allowing a cost avoidance of nearly 23.5 man-years. Of those cases, several were highly transactional, making them natural candidates for RPA. After reviewing and vetting the use cases, JMC decided to deploy digital workers into the program change request process within the Planning, Budgeting and Execution (PBE) system and Joint Reconciliation Process. Both processes were previously conducted by budget analysts within JMC’s G8 – Resource Management office.

“I am excited about the possibilities that process automation can bring to our area,” states John Campbell, Deputy Chief of Staff for RM. “We are using digital workers to research and compare financial profiles against future production schedules to ensure we are identifying any areas we should research further to maximize the customers ammunition buying power. This directly supports DOD’s audit of our financial statement as well as increasing the quality and efficiency of our work efforts to support the customer.”

Starting with RM made sense for JMC’s RPA deployment since many of its processes are highly transactional, static and less prone to change. JMC wanted to deploy a digital worker that would not require immediate restraint, especially during the developmental stage of a change management process for RPA. That change management process will be crucial in defining how to train digital workers before and after deployment. This will reassure enterprise system owners and partners that the digital workers are only performing functions they are trained to perform, which is an important consideration as JMC works to bring its initial RPA Proof of Concept into production.

JMC’s Proof of Concept focused on data fill errors that occurred during a data “handshake” between the Logistics Modernization Program and Munitions History Program that result in certain data fields being left empty. This left affected stock non-issuable until an analyst manually resolved the issue by filling in blank fields. This issue is common, with approximately 1,500 errors requiring mitigation each month. Due to resource constraints, not all errors get addressed, resulting in a backlog of approximately 50,000 unmitigated errors.

“All of the important factors were there to make the LMP/MHP data errors a priority use case for RPA” says Marc Dalmasso, JMC’s chief of Systems and Sustainment. “These errors cause a mission impact and take significant manpower to correct. However, LMP is an enterprise system, so we have to ensure we are closely coordinating with the system owner in order to gain the approvals needed to use RPA to access data.” That system owner is the Program Executive Office for Enterprise Information Systems, which defines LMP as “an Enterprise Resource Planning System that builds, sustains and generates warfighting capabilities using one of the largest, fully-integrated supply chain and maintenance, repair and overhaul solutions in the world.”

Given the magnitude and reach of LMP, JMC understood the need to bring the program lead to the table sooner rather than later to give a digital worker access to the system. JMC subsequently worked closely with personnel from the Product Manager, Logistics Modernization Program and Army Shared Services Center, which provides sustainment for the LMP system. As a result of these efforts, the parties outlined a course of action for testing and evaluation that, if successful, will result in the digital worker being granted access to LMP.

“That would be crucial,” said Dalmasso. “If we can begin accessing the ERP systems with digital workers, the sky is the limit as far as developing applications that will improve efficiencies and effectiveness at our command.” Importantly, effective application of RPA will improve JMC’s auditability and accountability, as well as improve the speed at which transactions can be processed. “This can have an immensely positive impact upon Army readiness in certain areas, says Crossen.

While JMC has primarily focused on RPA applications to this point, it also has a firm grasp on the importance of developing machine learning tools to help drive improved decision making for Army senior leaders. To that end, JMC has closely synchronized its RPA efforts with major players in the development of Machine Learning for the Army, Army Vantage and the Army Logistics Data Analysis Center. By bringing its AI requirements to the table early, JMC feels confident that it is prepared to jump onboard Army Machine Learning initiatives when the infrastructure becomes available for the establishment of an Enterprise data lake.

While machine learning gets most of the attention in the world of AI, JMC believes RPA is the foundation for future machine learning successes. “In fact,’ says Dalmasso, “RPA will play a critical role with machine learning. Our RPA efforts have demonstrated that digital workers can be used to improve data accuracy through elimination of human errors, but of equal importance is how they can be used to standardize processes.”

During its initial use-case analysis, JMC found that the data entry performed by human workers is subject to both error and ambiguity. Human workers often input data slightly different from each other, creating variability that makes data interpretation more difficult. “When data isn’t standardized, an analyst has to interpret the previous entries which not only introduces the potential for error, it takes time,” says Crossen. “If we can use digital workers to standardize the data, it ensures the data is accurate and timely. That will, in turn, improve the reliability and capabilities of our common operating pictures and the decision making tools we can develop using machine learning.”

While JMC has achieved several successes with RPA deployments, it is far from done. Over the course of 2020, JMC will expand its RPA footprint to include processes with sales orders and global freight management. JMC will also embark upon a combined RPA and machine learning effort for its Munitions Items Disposal Action System, which is used to characterize items requiring future demilitarization and disposal.

Expanding RPA and machine learning use cases requires expanding the skillsets and technical infrastructure needed to ensure continued AI successes. In order to address the need for internal expertise, JMC has begun training its workforce on AI and has an organic core of programmers excited to begin applying a new skillset. “It’s getting a lot of people excited,” says Dalmasso of possible AI applications for JMC, “it’s a new frontier and JMC wants to lead the way for the Army.”

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s