How the Pentagon’s JAIC Picks Its Artificial Intelligence-Driven Projects

The Pentagon launched its Joint Artificial Intelligence Center in 2018 to strategically unify and accelerate AI applications across the nation’s defense and military enterprise. Insiders at the center have now spent about nine months executing that defense driven AI-support. 

At an ACT-IAC forum in Washington Wednesday, Rachael Martin, the JAIC’s mission chief of Intelligent Business Automation Augmentation and Analytics, highlighted insiders’ early approach to automation and innovation. 

“Our mission is to transform the [Defense] business process through AI technologies, to improve efficiency and accuracy—but really to do all those things so that we can improve our overall warfighter support,” Martin said. 

Within her specific mission area, Martin and the team explore and develop automated applications that support a range of efforts across the Pentagon, such as business administration, human capital management, acquisitions, finance and budget training, and beyond. Because the enterprise is vast, the center is selective in determining the projects and programs best fit to be taken under its wing. 

“For the JAIC, there are a couple of key principles that we want to go by, or that we're going to adhere to when we're looking at…a project and whether we support it,” Martin explained.  

The first principle to be evaluated is mission impact. In this review, insiders pose questions like “who cares?” she said. They assess the user-base that would most benefit from the project and what the ultimate outcome would be across Defense if the JAIC opted to support it. Next, according to Martin, officials review data-readiness. In this light, insiders address factors like where the data to be used is stored—and whether it’s actually prepped for AI, or more advanced analysis and modeling to run on top of it. 

The third factor that’s assessed is technology maturity. Martin said that “contrary” to what many seem to think, the JAIC is not a research organization but instead seeks to apply and accelerate their adoption of already-existing solutions across the department and where those improvements are needed most. Insiders are therefore not at all interested in spending heaps of time researching new, emerging AI and automation applications. Instead, they aim to identify “what already exists and is ready to be deployed at this moment.” 

“So that's a big one for us that we like to emphasize,” Martin said. 

Related: Pentagon, NSA Prepare to Train AI-Powered Cyber Defenses

Related: Pentagon’s AI Center is Developing Tech that Could Revolutionize Disaster Response

Related: How AI Will Predict Chinese and Russian Moves in the Pacific

The final assessment is whether the JAIC can identify Defense insiders who will actually use whatever they are set to build. When developing something new, Martin said insiders want to those it’ll eventually touch to weigh in on the development every step of the way. 

“We're not in the business of coming up with good ideas and then creating something and trying to hoist it on somebody else,” Martin said. “We really believe in a very user-centric approach.”