The most surprising figure in the Defense Department’s latest AI update is not the drone budget, the procurement target, or even the rhetoric around lethality, but the AI adoption curve.
Emil Michael, the undersecretary of war for research and engineering and the department’s chief technology officer, declared that the use of AI across the department has increased by roughly 1.42 million users over the past year. That represents a 1,775% rise, increasing the number of AI users from about 80,000 to around 1.5 million across a workforce of more than 3 million personnel (US Department of Defense, 2026).
For years, military AI was often discussed in terms of prototypes, pilot programs, test environments, and future war concepts. Michael’s numbers suggest a different phase has begun. AI is being pushed into the working machinery of one of the world’s largest military organizations. “Every month, we have a new sort of amazing advancement in AI.” The real challenge, he added, is translating that pace “down to the Department of War and to the war fighter.”
The Military AI Stack Now Has Three Layers
Michael described the department’s use of AI across three dimensions: enterprise, intelligence, and warfighting. The enterprise layer is the least dramatic, but it may deliver some of the earliest gains. AI can reduce administrative drag, speed up internal processes, support procurement, improve logistics, assist maintenance planning, and help personnel navigate a vast bureaucracy.
The intelligence layer is more familiar. Modern militaries collect more data than analysts can process at human speed. Satellite imagery, drone feeds, signals intelligence, open-source information, and battlefield sensors generate a constant flow of material. AI can help sort, classify, summarize, and connect that information faster than conventional workflows allow.
The warfighting layer is the most consequential. This is where AI moves from assistance to combat relevance. Michael said the department is embedding AI into systems so war fighters can be “more precise,” “faster” and able to “make better decisions.” He also tied AI to shortening conflict, protecting personnel, and being “as lethal as possible.”
That language is stark, but it is also honest. In a military context, AI is being adopted because speed, precision, and decision advantage can determine battlefield outcomes.
Drone Dominance is the Visible Edge of the Strategy
The clearest example is the department’s Drone Dominance Program, which has allocated $1.1 billion to purchase 200,000 small lethal drones by 2027. That target says a great deal about where military technology is heading. Small drones sit at the intersection of autonomy, sensors, software, targeting, communications and mass production. They can scout, strike, confuse, overwhelm, and support operators in environments where larger systems may be too expensive, too visible, or too slow to adapt.
The goal of 200,000 drones points to a preference for scale and attainability. The future battlefield will not be shaped only by exquisite platforms that take years to build. It will also be shaped by cheaper systems that can be deployed in volume, upgraded quickly, and replaced when lost.
Michael explained that the department had previously relied on a limited group of approved drone vendors, a structure that gave companies little incentive to upgrade aggressively. His assessment was pretty blunt: “We kind of blew all that up.”
The new model aims to open competition, attract more vendors, and provide emerging companies a path to become prime suppliers of drone and counter-drone capabilities. That is not bureaucratic housekeeping. It is a recognition that slow acquisition can become a battlefield weakness.
Procurement Speed Is Now Part Of Military Power
For decades, defense acquisition was built around long development timelines, major platforms, and highly controlled procurement cycles. AI and autonomous systems move on a different clock.
Software improves quickly. Sensors change. Electronic warfare adapts. Counter-drone technology forces new tactics. A system that takes years to approve may arrive on a battlefield that has already evolved past it.
The department’s ability to compete will increasingly depend on how fast it can test, buy, field, and update technologies that improve through iteration. The old defense industrial model will not disappear, but it is being stretched. Traditional primes remain important, yet the AI and autonomy era requires a wider ecosystem: startups, software companies, robotics firms, data infrastructure providers, venture-backed manufacturers, and specialist engineering teams.
Rapid Adoption Brings a Governance Burden
Moving from 80,000 users to 1.5 million in a year raises difficult questions about training, access controls, data protection, model reliability, auditability, and human review. The stakes vary by use case. A flawed AI output in an administrative process may create inefficiency. A flawed output in intelligence analysis may distort priorities. A flawed output in a warfighting system can carry far more serious consequences.
The department’s AI surge cannot be judged only by adoption numbers. The deeper test is whether it can build guardrails at the same speed it builds capability. Military AI needs clear rules around when systems can recommend, when they can act, and when human judgment must remain decisive. It also needs testing environments that reflect real operational conditions, because battlefield data is rarely clean, complete, or stable.
Special Operations as a Natural Proving Ground
It is quite noteworthy that these remarks were made at a special operations event. Special operations forces have historically been known for their agility, adaptability, small team sizes, unconventional missions, and close integration with emerging technologies. That makes them a logical early audience for AI-enabled tools. A small unit operating under pressure may benefit from faster intelligence synthesis, better drone integration, improved targeting support, automated translation, logistics assistance, route planning, or real-time sensor fusion.
But special operations also highlight the limits of the technology. These missions often depend on judgment, context, cultural understanding, ambiguity, and restraint. AI can support those qualities, but it cannot replace them. The strongest use case is not machine autonomy detached from human command. It is better human decision-making under conditions where time, information, and risk are all compressed.
The Real Struggle… Operational Integration
Michael ended with a message to special operations forces, saying they are the “heart and soul” of what the department does and that he wakes up thinking about how the department can better serve them. The department is racing toward models, drones, and autonomy, but the measure of success will be whether those tools actually help operators in the field.
But can a vast military bureaucracy absorb fast-moving technology without becoming slower, riskier, or more confused? The answer will depend on more than algorithms. It will depend on procurement reform, technical talent, disciplined governance, operator trust, and the ability to turn innovation into usable capability.
AI is becoming the Pentagon’s new operating layer but whether it becomes a durable advantage will depend on how intelligently the department manages the space between speed and control.