Modernized Acquisitions Accelerate DOD AI Development

Modernized Acquisitions Accelerate DOD AI Development

The agency's contracting and development process reforms lead to quick turnarounds for artificial intelligence capacities for the American warfighter.

The Defense Department's Joint Artificial Intelligence Center is exploring a modernized approach to contracting that accelerates its rate of technological adoption and more efficiently delivers the benefits of AI capacities to the U.S. Armed Forces as a whole.

With emerging technologies taking hold at the department, it faced a challenge in devising a modernized solution amid the DOD's often conservative approach to change, JAIC Acquisitions Chief Will Roberts told GovernmentCIO Media & Research.

“What we've found more and more throughout the contracting process … is that for us to deliver AI capabilities, we need to modernize many aspects of the way we do things. Including even HR, how we protect our security systems, but especially acquisition,” Roberts said.

In the past, JAIC has explored integrating more automation in the acquisition process itself to speed up lead time.

Other means of finding ways to keep pace with technological change allowed products to be iterated as they’re being developed while demonstrating their concrete benefits to DOD leadership.

“The need to acquire things at the speed of relevance is a big theme," he said. "It is with every emerging technology, but especially with AI. We've come to find out that the technological advances in our marketplace used to be once every 10 years or so. Now that's greatly shortened to where we have huge impacts on technology within the next year or so."

As a consequence, relying on traditional fixed acquisition styles would prove far less efficient for delivering on the benefits of AI to the American warfighter. This has led JAIC to take an iterative approach to AI development that keeps a more open relationship with industry partners while allowing DOD technologists to evaluate certain products beforehand.

“We are finding out that we can create these agreements where the government pays by either giving [industry] access to DOD data for a temporary time, or we pay by giving them a testing report. And the testing report will outline how well the product did,” Roberts said.

Roberts also outlined how this has required nurturing industry savvy among DOD contracting officials so as to better understand how to engage with the rapidly changing nature of artificial intelligence development and integrate this within the product life cycle.

“The DOD needs to understand that their business acumen needs to go off the charts. So it's not just that they're filling out a market research report or going through the contracting process with due diligence and finding just two or three companies. They need to have their finger on the pulse of the AI industry as a whole. They really need to know what's out there, and they need to engage in this transparent feedback with industry,” Roberts said.

This foundational practice has made it easier to engage in agile methodologies rapid prototyping that can facilitate the product development cycle while allowing testing and calibration that better keeps pace with the rate of technological change.

“Once they have a good statement of objectives through robust collaboration with industry, they can move into a repeatable predictable and rapid process to get a prototype out the door,” Roberts said.

Despite the newness and complexity of breaking AI developments, JAIC’s modernization push ultimately hinges on using these novel technologies to support the DOD’s foundational mission and bringing their benefits in a user-friendly way to the warfighter themselves.

“What's even more essential is that the end user who's going to be taking on these capabilities might be someone who doesn't understand AI, nor do they need to. User interface is a huge piece of that — getting the user interface right, getting that feedback loop right with the soldiers in the field. Making sure we have that iteration done is a big piece of JAIC’s mission. We can't stress enough this idea of a customer-friendly, user-friendly process,” Roberts said.

Standard