Everyone is talking about artificial intelligence, but actual no-kidding military applications can be hard to identify.
“If you have a data problem, or if you can make a problem into a data problem, it’s probably a good fit for AI,” says Angela Sheffield, an internationally recognized expert in nuclear nonproliferation and applications of AI for national security.
Sheffield has been cited for “transforming” the National Nuclear Security Agency’s Office of Defense Nuclear Nonproliferation with innovative AI research and development. There, as Senior Program Manager for AI and Data Science, Sheffield developed next-generation tools for detecting early indicators of illicit nuclear weapons development.
Now Sheffield has a new role as director of AI programs at SAIC, a leading systems integrator and solutions provider for federal and defense applications. A former Air Force intelligence officer, she sees numerous opportunities to bring AI to a host of defense requirements and says getting started is often the hardest part, because it means getting past all the reasons not to move forward.
“We will forever have legacy systems,” Sheffield says. “We will always have fragmented and siloed data repositories. Those aren’t things that we can wish away.”
But they also don’t need to be barriers to automation. Whether one is tackling a complex problem like Combined Joint All Domain Command and Control, major initiatives to modernize weapon systems, or efforts to automate Tasking, Collection, Processing, Exploitation, and Dissemination (TCPED), mundane tasks that involve routine work can be automated to to reduce the human cognitive load.
“There are a lot of other applications ripe for opportunity, for modernization and innovation and AI,” Sheffield said. And it doesn’t have to be the hard, super complex use cases: Business operations and other routine operational tasks “are really great opportunities for us to leverage AI. Automation can free up our Airmen and our other service members and civilians … to tackle the challenges DOD faces today.”
In some ways, that is beginning to happen. Enterprise IT is gaining a foothold with AI-driven capabilities integrated into email and other collaboration activities.
“We’re beginning to expect that as the part of the services that we get from our enterprise IT,” Sheffield said. AI can also support efficiencies, she added, in business operations and mission execution “to fulfill requirements in computing, in managing disparate data sources.”
This is where an integrator can be especially valuable. “SAIC is part of bringing those solutions to the Air Force and the rest of the joint force, with concepts like data layers that interconnect stove-piped or fragmented data systems,” she said.
Once data can be shared across systems, everyone benefits: “You can get a single-site picture or a single understanding of all of your resources captured in those different repositories,” she said, enabling AI-supported process automation, enhanced analytics, and informed, accelerated decision-making.
Users will not necessarily buy into automation easily, she said. Trust must be earned—and built—over time to ensure users gain confidence in intelligent systems. They need to see that the software works, Sheffield said, and “to understand how it is working, if it’s performing within the intended envelope.” And they need to be confident that the AI is not generating erroneous “hallucinations,” she said.
AI must be a primary driver for enabling CJADC2 because without it the data sets are too large, the problems too demanding, to maintain an information advantage at the speed of modern warfare. CJADC2 demands real-time sharing of data across service, national, and digital boundaries.
That means overcoming legacy IT roadblocks and information systems that can’t talk to each other. Interoperable databases and AI-driven automation are part of the solution. “CJADC2 will happen as a result of that modernization in a way that’s even more powerful than what we’re beginning to see in pilot demonstrations,” Sheffield predicts.
For example, Indo-Pacific Command’s Joint Fires Network, a Battle Management System delivers real-time actionable threat data to joint, partner, and allied forces. SAIC is involved in that pilot, and Sheffield foresees more AI-driven implementations like it, “where we’re closing kill chains faster and achieving those successes.”
Disparate systems and technologies, often purpose-built with proprietary technology, must be integrated to make them work. “That’s where an integrator like SAIC can help,” she said.
As a federal program manager, she recalled, “I often relied on my contractors or performers to provide that visibility — lessons learned from one agency to another,” she said. Commercial partners “helped me have that visibility of what’s happening across the interagency.”
That’s exactly the value Sheffield says she brings to her work at SAIC. “Looking across our multi-mission portfolio and bringing the best solutions for DOD’s missions is something they can rely on us to do,” she said.