Acquiring Artificial Intelligence Systems: Development Challenges, Implementation Risks, and Cost/Benefits Opportunities

Abstract

The acquisition of artificial intelligence (AI) systems is a relatively new challenge for the U.S. Department of Defense (DoD). Given the potential for high-risk failures of AI system acquisitions, it is critical for the acquisition community to examine new analytical and decision-making approaches to managing the acquisition of these systems in addition to the existing approaches (i.e., Earned Value Management, or EVM). Also, many of these systems reside in small start-up or relatively immature system development companies, further clouding the acquisition process due to their unique business processes when compared to the large defense contractors. This can lead to limited access to data, information, and processes that are required in the standard DoD acquisition approach (i.e., the 5000 series). The well-known recurring problems in acquiring information technology automation within the DoD will likely be exacerbated in acquiring complex and risky AI systems. Therefore, more robust, agile, and analytically driven acquisition methodologies will be required to help avoid costly disasters in acquiring these kinds of systems. This research identifies, reviews, and proposes advanced quantitative, analytically based methods within the integrated risk management (IRM) and knowledge value added (KVA) methodologies to complement the current EVM approach.