The Pentagon’s push to integrate emerging technologies in its financial management and budgeting process has resulted in finding $4 billion in dormant contract obligations to shift toward higher priority items and helped prevent millions of dollars in improper payments, an official said Tuesday.
Gregory Little, the department’s deputy comptroller for Enterprise Data and Business Performance, said his office is specifically looking at further opportunities to partner on projects related to artificial intelligence, machine learning, cloud computing, software-as-a-service, data analytics, robotic process automation, natural language processing and blockchain technologies.
“I think financial management has really embraced digital transformation innovation,” Little said during a
FedInsider webinar discussion. “We’ve seen a lot of successes and we’ve learned a lot. We’ve also started embracing an agile approach where we build a little, test a little and learn a lot. I think we recognize that there’s an opportunity cost if we don’t move quickly.”
Little said the work to find $4 billion in dormant contract obligations was the result of partnering with several DoD offices on AI and machine learning tools paired with basic descriptive analytics, while a new application was created to help highlight the millions in improper payments.
“If we don’t see any activity in those obligations, and we have a sampling and a risk methodology, we alert the users to those. Then those users go in and, if that transaction is not needed anymore with that contract and with the remaining funds, we can actually deobligate those funds and apply those funds to more high priority items,” Little said. “We’re actually doing it in a cheaper way by leveraging better technology and better analytics.”
In February, the Government Accountability Office’s director of defense capabilities and management told lawmakers the Pentagon estimated it paid about $11.4 billion in improper payments in fiscal year 2020 alone.
Little said his office’s work on natural language processing has included deploying algorithms to free up 30,000 hours in labor from predictive maintenance to more intensive tasks and as well partnering with the Joint Artificial Intelligence Center on a project to streamline new policy management.
“We’ve actually sucked all of the department’s policy data into this platform, all of those PDFs, so we can actually establish relationships with other policies. Now, when our policymakers are looking for duplicative policy or how to streamline policy or what are the impacts of making a policy update, we’ve created some really easy tools for them to be able to do that with some really interesting, complex technology on the backend,” Little said.