The Pentagon is racing to deploy artificial intelligence across military operations, but structural barriers rooted in decades of bureaucratic tradition, industrial decay, and technical complexity may prevent the rapid transformation that military leaders say is essential to maintain American military superiority. A new Carnegie Endowment analysis identifies eight specific bottlenecks that will slow AI diffusion in the U.S. military, using autonomous drones as a case study for the broader challenge.

The urgency is real. AI-enabled command-and-control systems and semi-autonomous drones are already reshaping battlefields in Ukraine and the Middle East, while China is mobilizing its industrial base to establish military AI dominance. The White House has called for the military to "aggressively adopt" AI, and Secretary of Defense Pete Hegseth's strategy declares that the Department of Defense (DoD) must become an "AI-first" warfighting force. But the gap between ambition and execution is widening.

What AI Is Actually Doing in Today's Military?

AI is already playing a measurable role in U.S. military operations, though mostly in narrow, decision-support applications rather than autonomous systems. Palantir's Maven Smart System, which uses computer vision and large language models (LLMs) like Anthropic's Claude, has accelerated targeting operations in Iran. The system increased targeting speed from fewer than 100 targets per day to 1,000 per day using computer vision alone. When Claude was integrated into Maven's workflow, that figure jumped to 5,000 targets per day.

The U.S. military struck more than 13,000 targets during the first thirty-eight days of the Iran conflict using AI-assisted targeting. Claude also reportedly played a role in the operation to capture Venezuelan President Nicolás Maduro, though the specifics remain classified.

However, speed in processing targets does not equal strategic success. The massive scale of aerial bombardment in Iran has struggled to produce favorable strategic outcomes, much like the hundreds of thousands of attack sorties in Vietnam. Processing data faster is fundamentally different from the kind of transformational change that true military AI adoption would require.

What Are the Eight Barriers Slowing Military AI Adoption?

The Carnegie analysis identifies a range of interconnected obstacles that will impede AI diffusion across the military, even under pressure. These barriers span cultural, bureaucratic, technical, and industrial dimensions:

• Entrenched Cultural Norms: Decades of military preeminence have created institutional traditions and hierarchies resistant to rapid technological change and risk-taking.
• Slow Political and Bureaucratic Processes: Decision-making timelines in the Pentagon move at a pace incompatible with the speed of AI development and deployment in the private sector.
• Degraded Industrial Base: The U.S. defense manufacturing ecosystem has atrophied, limiting the ability to scale production of autonomous systems quickly.
• Supply Chain Vulnerabilities: Producing autonomous drones en masse would require alternatives to Chinese components that do not yet exist.
• Testing and Evaluation Gaps: Modernizing the Pentagon's testing capacity and red-teaming environments is necessary but time-consuming.
• Doctrine and Force Structure Misalignment: Using autonomous systems effectively would require redesigning military doctrine, force structures, training regimens, and logistical chains.
• Talent Recruitment and Retention: The military struggles to attract and retain technically skilled personnel who can develop, field, and adopt AI tools.
• Integration Complexity: Scaling AI tools across diverse military platforms and missions is highly complex, regardless of urgency or resources.

Would a Major War Change Everything?

An existential threat from a peer adversary like China could reduce most barriers to AI adoption. Total war would unlock political will for intensified research and development, streamlined regulatory reviews, and broad buy-in at every military level to adopt promising AI tools. However, even abundant resources and extreme urgency cannot solve all problems overnight.

The analysis warns that some technical and industrial challenges require time that a major conflict may not provide. Producing autonomous drones at scale, redesigning military doctrine, and integrating new systems across the force are complex undertakings that cannot be compressed into weeks or months, even under wartime pressure. By the time a peer conflict begins, it may be too late to complete the necessary AI transformation.

How Should the Pentagon Accelerate AI Adoption Now?

The Carnegie analysis recommends a series of concrete steps the Department of Defense should take immediately to build AI capability before a crisis forces the issue:

• Expand AI Talent Pipelines: Recruit and retain technically skilled personnel not just in headquarters but as operational warfighters, and establish feedback mechanisms that reward candor and experimentation.
• Incentivize Industrial and Operational Risk: Invest in small, specialized firms, including foreign companies, and quickly discard failures to accelerate production and experimentation cycles.
• Modernize Testing and Oversight: Invest in red-teaming and testing environments, establish clear minimum requirements for capability development and cross-systems integration, and preserve meaningful human authority over lethal decisions.
• Treat Autonomous Drones as a Pacing Challenge: Use autonomous drones as a benchmark for wider AI transformation, since they represent some of the hardest technical and integration problems the military will face.

The analysis emphasizes that testing, evaluation, and oversight should be modernized, not eliminated. Gutting the Pentagon's testing capacity or carelessly offloading lethal decisions to autonomous systems could endanger soldiers' lives and lower the threshold for conflict. Trust in AI systems will ultimately determine the scale of their use.

Autonomous drones are just one application of military AI, but the challenges they face generalize across the broader AI adoption agenda. Software-only AI tools are simpler to build and scale than physical autonomous systems, but they too will encounter technical limits, institutional inertia, and integration hurdles. The barriers identified in the drone case study apply to decision-support systems, logistics optimization, and other AI applications the military plans to deploy.

The Pentagon's AI transformation is not a question of technology alone. It is a question of whether the U.S. military can overcome decades of institutional inertia, rebuild its industrial base, and cultivate a culture of rapid experimentation before a peer adversary forces the issue in combat.