DOD, Army looking to streamline, fast-track AI acquisition; but it’s slow

The U.S. Department of Defense (DOD) hopes that its new Open Data and Applications Government-Owned Interoperable Repositories (Open DAGIR) Challenge initiative will streamline how the US military acquires AI and to scale it across the services.

The U.S. Army, meanwhile, is looking at establishing a separate sub-path within its larger software acquisition program specifically for AI.

The Pentagon has been working in earnest to develop AI capabilities that can do tasks that normally require human intellect.

Open DAGIR was launched in July by the Chief Digital and Artificial Intelligence Office (CDAO) in partnership with the Office of Under Secretary of Defense for Research and Engineering, the Joint Fires Network, and the Defense Innovation Unit.

The CDAO said at the time that it “remains focused on driving a healthy, competitive landscape for data, analytics, and AI platforms, tools, and services, and that it “will allow DOD to match the pacing challenge, incentivize innovation, and drive industry to offer its best technical solutions to address critical mission needs.

In late August, the CDAO announced that Open DAGIR was being hosted in the Tradewinds Solutions Marketplace and will support and combine a highly innovative acquisition process with the department’s Combined Joint All Domain Command & Control (CJADC2).

DOD’s Tradewinds Solutions Marketplace enables industry, academia, and other innovators to get their solutions in front of government buyers. It is a venue for DOD organizations to search, view, review, compare, contrast, contact, discuss, negotiate, and procure data, analytics, digital, and AI/ML capabilities solutions through rapid acquisition procedures.

The Open DAGIR Challenge is a call for vendors interested in submitting solutions in the contested logistics and sustainment space to support the CJADC2 effort. DOD says the Challenge will transform the Global Information Dominance Experiments (GIDE) by leveraging the entire industrial base to rapidly apply industry solutions to warfighter needs.

“The initiative has the potential to impact every combatant command, and warfighters will witness industry solutions applied immediately to their problem sets in a common CJADC2 global integration decision platform,” DOD said, adding that “warfighters will rapidly share feedback in an iterative fashion, while executing the mission.”

“We are thrilled about the recent adoption of Open DAGIR, which enables the department to incentivize innovation by driving a competitive environment for data, analytics, and AI platforms, tools, and services,” US Air Force Col. Matthew Strohmeyer, GIDE Mission Commander said.

The Open DAGIR program establishes a multi-vendor ecosystem to enable industry and the government to integrate data platforms, development tools, services, and applications, while preserving the government’s data ownership and industry’s intellectual property, something the contracting community had expressed concerns about to the Government Accountability Office (GAO).

GAO noted in a June 2023 audit report that 13 private companies had informed it of “the importance of considering intellectual property and data rights when negotiating contracts [with DOD] for AI projects.”

GAO said while “parts of DOD are already using AI, DOD hasn’t issued department-wide AI acquisitions guidance needed to ensure consistency.” GAO recommended at the time that “it develop such guidance – considering private company practices as appropriate.”

DOD has designated AI as a top modernization area and is allocating considerable spending to develop AI tools and capabilities, everything from identifying potential threats to targets on the battlefield.

At the time of GAO’s audit, it said that while “numerous entities across DOD are acquiring, developing, or already using AI,” DOD had not issued department-wide guidance for how its components should approach acquiring AI [and] had “not defined concrete plans and has no timeline to do so.”

GAO also stated that “the military services also lack AI acquisition-specific guidance, though military officials noted that such guidance would be helpful to navigate the AI acquisition process. Without department-wide and tailored service-level guidance, DOD is missing an opportunity to ensure that it is consistently acquiring AI capabilities in a manner that accounts for the unique challenges associated with AI.”

GAO added that various DOD components and military services had individually developed or planned to develop their own informal AI acquisition resources, and that some of these resources reflected “key factors identified by private companies for AI acquisition.”

“For example,” GAO noted, “DOD’s Chief Digital and AI Officer oversees an AI marketplace known as Tradewind, which is designed to expedite the procurement of AI capabilities,” and that “several Tradewind resources emphasize the need to consider intellectual property and data rights concerns when negotiating contracts for AI capabilities, a key factor identified by the companies GAO interviewed.”

Deputy Secretary of Defense Kathleen Hicks in November 2023 announced DOD’s Data, Analytics, and AI Adoption Strategy developed by the CDAO. It unified previous strategic guidance to scale advanced capabilities across the services.

The first DOD AI Strategy, published in 2018, and a revised DOD Data Strategy published in 2020, are the two foundational documents that matured the Pentagon’s data-centric structures to increase the efficacy of fielding modern AI-enabled capabilities.

“We’ve worked tirelessly, for over a decade, to be a global leader in the fast and responsible development and use of AI technologies in the military sphere, creating policies appropriate for their specific uses,” Hicks said at the time,” pointing out that, “as we’ve focused on integrating AI into our operations responsibly and at speed, our main reason for doing so has been straightforward: because it gives us even better decision advantage than we already have today.”

About five years ago the Pentagon created six acquisition pathways, including a software acquisition pathway, that was designed to streamline and fast-track development and delivery of AI and AI-enhanced software. Implementing the policy, however, has been sluggish. It has been reported that only about 50 programs across the services have or are utilizing the software pathway. The Army has only used the pathway for 19 of its more than 500 programs.

Young Bang, Principal Deputy Assistant Secretary of the Army for Acquisition, Logistics & Technology, told Federal News Network that the Office of the Secretary of Defense is working “to get to something faster for AI because as fast as the software pathway is, we need a faster path for algorithms.”

Bang said, “We’re experimenting and feeding and working with them to figure out a subpath that will be the only one right now. Previously, it was the only one with two pathways. This potentially may be three or will inform the subpath, but that’s important.”

Bang said the Army is currently working with Deborah Rosenblum, the acting deputy undersecretary of defense for acquisition and sustainment, on whether there should be a subpath within the software pathway specifically for AI.

Rosenblum has said that the software pathway is likely to be the best option because this pathway already supports rapid development, prototyping, and fielding.

“I do believe it’s likely to be through the software pathway, if only because we’re trying to expedite and move things quickly. And we feel that the way in which the pathway is designed has enough flexibility and agility,” Rosenblum said.

Under DOD’s existing software pathway, programs must reach a Minimum Viable Capability Release (MVCR) within one year.

A MVCR is designed to provide minimum capability in an operational environment to provide value to the end user or warfighter in a short amount of time.

“If you think about what a minimal viable capability release is, it takes a year to get there. I can train algorithms literally overnight,” Bang explained to Federal News Network.

“What we’re saying is there’s great utility in a software pathway, but if we use a software pathway for algorithms — overnight is a good example, but some of these can actually take a little bit longer, but still a week. And if we think about that, a week versus an MVCR in a year — the timelines don’t align.”

“Once you make the algorithms, you can extend and retrain them at the edge. You can do inverse learning, and you can do that quickly. And if we’re governed by a year-long MVRC process, we don’t think that’s fast enough to get capabilities out to our soldiers,” Bang said.

“My ultimate goal is to get capability to the warfighter as quickly as we can,” Plumb said in a recent interview with Dcode Co-Founder and CEO Meagan Metzger. “A key part of that is not just being clear on the pathways and requirements, but creating some transparency on that pathway, so that we can do kind of leverage the secret sauce of American investment.”

Article Topics

AI  |  biometrics  |  Department of Defense  |  government purchasing  |  military  |  U.S. Army