Human-AI Teaming Under Fire: Lessons from Ukraine's Human-in-the-Loop Combat AI Systems

Abstract

This article examines human-artificial intelligence (AI) teaming in Ukrainian combat operations from 2022 to 2025, exploring the integration of AI systems with human decision-making in military contexts and how crisis-driven innovation can lead to human-AI teaming. The research addresses three questions: the effectiveness of human-AI teams compared with human-only or fully autonomous systems; the effect of organizational structures on the sustainability of AI integration; and strategic implications for the development of the doctrine and international security governance in the case of the North Atlantic Treaty Organization (NATO). The methodology uses mixed-method comparative case study analysis of three different Ukrainian systems: the Geographic Information System (GIS) Arta geospatial intelligence platform, the reconnaissance-strike unmanned aerial vehicle complex, and volunteer-supported decentralized targeting networks. Data collection was a combination of technical reports, operational battlefield metrics, and the standards of the NATO doctrines. Findings show that Ukrainian human-AI systems show good tactical performance in permissive electromagnetic environments with response times of 30 to 45 seconds and targeting accuracy of two meters but have significant vulnerabilities to electronic warfare—31% mission failure rates. Volunteer networks are highly resilient and have slower decision cycles. The research adds to strategic security, deterrence theory, military innovation theory, and organizational theory through the identification of mechanisms by which human-AI systems affect the stability of deterrence and offers recommendations for the development of the doctrine and international governance of AI for NATO.