Enhancing Decision Accuracy in DoD Acquisition: Integrating Artificial Intelligence with Reference Class Forecasting

Abstract

"This capstone examines how integrating artificial intelligence (AI) with reference class forecasting (RCF) can improve decision accuracy in Department of Defense (DoD) acquisition. Persistent cost overruns and schedule delays, driven by optimism bias and planning fallacy, highlight the limits of traditional forecasting. These shortfalls routinely undermine mission readiness and erode fiscal discipline. While RCF enhances accuracy by anchoring estimates in historical data, its use in the DoD is limited by scalability and data-access challenges. This study uses a qualitative design combining policy review, comparative case analysis, and conceptual modeling. The findings indicate that AI can support the automation of reference-class construction from unstructured acquisition data and enable probabilistic forecasting, improving cost and schedule realism. Supported by prior literature and simulated analysis, the results also suggest that this approach can strengthen technology readiness assessments (TRA) by incorporating risk-based probability bands, thereby highlighting the value of probabilistic evidence in early acquisition decision-making. Recommendations include phased AI–RCF implementation, governance standards for transparency, and integration into milestone artifacts like TRAs and Life cycle Sustainment Plans. Institutionalizing this approach would embed empirical rigor into acquisition decisions, reduce systemic risk, and advance the DoD’s shift toward data-driven reform."