DoD's New 'Adopt, Buy, Create' AI Strategy is Built on Software and Quality of Data

(Marina Varnava/Adobe Stock)

developed by the Chief Digital and AI Office (CDAO). The 26-page document  outlines the agency's new enterprise-wide "Adopt, Buy, Create" approach to procuring artificial intelligence (AI) that favors application programmable interfaces (API), data analytics and software over new artificially intelligent hardware, weapons or kinetic systems.

Released several days after the White House published its own new AI executive order , this the latest AI guidance and strategy document released by DoD since its 2020 Data Strategy that was published two years after the 2018 AI Strategy and establishes a new "AI Hierarchy of Needs" with quality of data as a foundation. While CDAO expects to publish official implementation guidance within the next few months, DoD officials provided details about what it means for government, industry and university engineers working on AI, software and data analytics technologies for military assets and programs.

One of the major overarching goals of the strategy is to effectively establish a DoD-wide digital data feedback loop about in-service assets and services such as aircraft, satellite communications and warships among others. According to a section of the new strategy focused on AI adoption, the new "adopt-buy-create" framework is aligned with DoD's Software Modernization Strategy and the Office of Management and Budget (OMB) Circular A-130, "Managing Information as a Strategic Resource." The strategy segments DoD assets, technologies and programs into individual "Components" that will be responsible for assessing data across individual asset lifecycles using the data quality dimensions first outlined in the 2020 DoD Data Strategy.

Under this framework, individual defense procurement officials will "first seek to adopt solutions that are already Joint- or Component-sponsored before exploring capabilities available on the open market," the strategy says This will prioritize investment by the agency in new applications, software, data analytics and AI-enablement software that are already being developed as part of DoD sponsored competitions, prototyping or other efforts to develop new capabilities.

"When DoD-owned shared services are unavailable, the Department will challenge vendors to solve specific business and mission problems, while designing acquisition strategies to avoid vendor lock-in," DoD program officials write in the new strategy. "Fielding web-based, cloud-based, and/or Application Programming Interface first applications create more opportunities for rapid, enterprise scalability; continuous integration and delivery; and increased economies of scale. APIs also allow for a more open exchange with diverse data sources, regardless of origin."

(The hierarchy of AI needs as outlined in the new AI data analytics strategy document. Image: DoD)

Furthermore, DoD will only develop its own internal AI tools or data analytics technologies when those type of applications cannot be readily adopted from commercial or existing solutions. These DoD-specific applications often involve embedded software coupled to customized DoD hardware, the strategy notes.

The agency is also establishing standard language necessary for data, analytics and AI technology "contract problem statements and agreements.” To expedite acquisitions, DoD officials will use the federated data and model catalogue for AI-enabled enterprise tools. There is also an ongoing effort to increase government visibility over the ownership, labeling, maintenance and classification of DoD data.

During a media briefing about the new strategy, Deputy Secretary of Defense Kathleen Hicks said it will help the agency’s realization of the combined joint all-domain command and control (CJADC2) concept of warfighting operations. She also acknowledged aerospace and defense industry advancements in AI while noting the need for improvement to meet DoD’s ethics and responsibility requirements.

“Candidly, most commercially available systems enabled by large language models aren't yet technically mature enough to comply with our ethical A.I. principles, which is required for responsible operational use,” Hicks said. “But we have found over 180 instances where such generative A.I. tools could add value for us with oversight, like helping to debug and develop software faster, speeding analysis of battle damage assessments, and verifiably summarizing texts from both open source and classified data sets.”

DoD had already been experimenting with generative A.I. tools “before ChatGPT,” Hicks said, with some of those tools developing their own models, “isolating foundational models, fine-tuning them for specific tasks with clean, reliable, secure DOD data, and taking the time to further test and refine the tools.” Hicks also emphasized the need for ongoing human involvement with the deployment of new AI capabilities.

Chief Digital and AI Officer Craig Martell also participated in the media briefing and emphasized the strategy’s focus on AI-enabled software and data analytics over new AI-enabled hardware or weapons. Martell provided examples of some initial use cases, such as placing a “business health dashboard” on Hicks’ desk to demonstrate the type of AI-enabled data feedback loop the agency wants to establish about its assets, services and programs.

“This is not a capability development strategy. Technologies evolve, things are going to change next week, next year, and next decade, and what wins today might not win tomorrow,” Martell said. “Rather than identify a handful of A.I.-enabled warfighting capabilities that will beat our adversaries, our strategy outlines the approach to strengthening the organizational environment within which our people can continuously deploy data analytics and AI capabilities for decision advantage.”