The Pentagon’s AI Bet Is No Longer Theoretical
Strategy at the intersection of AI, defense resources, acquisition and execution
A portfolio operating model for speed, control and measurable mission value
Artificial intelligence is moving rapidly from experimentation into the Department’s operating environment. The harder challenge now is not adoption. It is building the financial, technical and decision architecture required to scale AI without losing accountability.In this independent strategy memorandum, Mayra G. Ugaz Rivera examines military AI through a portfolio, acquisition and financial-management lens.
The paper proposes a practical operating model centered on four controls:
Model Promotion Gate
Evaluate models by version, mission performance, cyber risk and consequence before production release.
Agent Identity + Authority
Give every persistent operational agent a named human owner, bounded permissions, expiration and an auditable activity record.
Open Model Gateway
Keep commercial models replaceable so the Department can adopt better capabilities without rebuilding mission workflows around a single vendor.
Investment-to-Outcome Lineage
Connect requirements, appropriations, contracts, obligations and expenditures to delivered capability, adoption and verified mission outcome.
The analysis also examines the FY27 AI and CJADC2 investment picture, sovereign compute, the War Data Platform, GenAI.mil, agentic AI, human judgment, testing, cybersecurity, full-stack cost and the financial-governance challenges that emerge as AI becomes embedded inside larger defense portfolios.
The central question
How much are we actually spending on AI, where is it funded, where is it executing, and what measurable mission outcome did each increment of investment produce?
The paper concludes with a proposed 90-day portfolio-control pilot and a framework for scaling only those AI capabilities that can be measured, governed and replaced.
Author: Mayra G. Ugaz Rivera
Published: August 2026
Type: Independent Strategy Memorandum
