Join War on the Rocks and gain access to content trusted by policymakers, military leaders, and strategic thinkers worldwide.
The military’s AI strategy is starting to resemble a bad dating app: lots of swipes, few lasting matches. The Trump administration’s new AI Action Plan offers potential to take things to the next level. It calls for deeper investments by the U.S. Department of Defense in algorithmic warfare. The plan calls for turning senior military colleges into hubs for AI research where students learn and apply core AI skills. This approach complements calls in the Army Transformation Initiative to integrate AI agents into next-generation command and control.
Realizing the Trump administration’s vision of algorithmic warfare will require radically rethinking what constitutes senior-level education in the profession of arms and embracing key ideas about creating data-driven learning organizations.
Indeed, senior military colleges should become centers for AI training, research, and application, with classrooms functioning more like battle labs than lecture halls. Professional military education should serve as an organic refinement ecosystem — a modern version of the general staff — where student and faculty problem-solving is captured and fed back into AI systems.
Teaching should focus less on general reflections on history and theory and more on solving pressing operational problems, with the best student solutions serving as an organic feedback loop that builds reinforcement learning with human feedback into the modern profession. These schools should become smaller but more elite, with fewer administrators and faculty who combine teaching with applied research on war, ensuring the classroom itself becomes a site where new knowledge is created and refined.
Done right, these schoolhouses could truly be the new brain of the military.
The State of AI in the Department of Defense
The generative AI turn is still an open market yet to be consolidated in the U.S. military. Foundation models such as ChatGPT, Llama, Gemini, and Claude, along with applications like Palantir’s Maven, Scale AI’s Donovan, and others, present themselves as the path to the future of planning and decision-making. Next to real progress are products polished for show but empty in substance. Hackathons, experiments like Thunderforge, and endless “military GPTs” blur the line between innovation and illusion, leaving leaders and staff officers struggling to separate what is useful from what is hype.
The spectacle distracts from the harder task of building the intellectual foundations that AI will need. Everyone poses as an AI expert, but few grasp the underlying technology well enough to apply it in war. The probability of believing the hype increases with every bad claim by industry, teachers, and peers alike. Students leave with unrealistic expectations or become numb, preferring to tune out rather than lean into building their own AI agents and narrow use-case applications.
What is missing is a deliberate effort to organize history, theory, and doctrine into usable forms that machines can learn from. Without this curated foundation, models will continue to know more about celebrities than about Carl von Clausewitz or William Slim. The military professional becomes lost in the noise.
If the military is serious about AI, it should prioritize expertise over performance. The real challenge is curating domain-specific knowledge that allows algorithms to deal with fog, friction, and shifting aims in war, and to apply concepts like operational design, centers of gravity, and emerging ideas such as pulsed operations and mosaic warfare.
Building the Brain of the Officer Corps
In an era of agentic AI, the military will need a collective mind, with officers able to think critically and challenge what algorithms produce. Professional forums should be places where learning happens alongside the creation of new knowledge that feeds directly into the systems used for planning and operations. Building this brain cannot be outsourced, since it depends on capturing how the military’s best thinkers approach problems and apply those insights to refine foundational models.
This brain of an army is an extension of an old idea — the general staff — but adapted to modern information networks. In the late 19th century, Spencer Wilkinson published The Brain of an Army, describing the German general staff and its key role in planning successful military campaigns. The Prussian model, with its distinct ideas about education and how to unleash “enlightened soldiers,” became the standard for military organizations. This model holds a special place for a mix of campaign analysis through historical cases and decision games, captured in Clausewitz’s On War in the idea of critical analysis in Book 2, Chapter 5. The true professional uses history to accumulate insights into the complexity of warfare, but with an applied eye toward future campaigns. Studying war and planning become co-constituted and inseparable, creating a need for less lectures and meandering seminar discussions and more time spent designing and fighting campaigns. That means civilian faculty will need to spend more time learning about joint all-domain operations and how to plan rather than hiding in disciplinary comfort zones.
Done right, this process produces new knowledge. Students have to do more than show they can solve stale “schoolhouse” solutions. They should demonstrate they can be thinking agents — neurons in the brain of the military, capable of generating estimates, analyzing courses of action, and functional planning. This analytical process hones their judgment while producing options for solving key operational problems and identifying requirements for building the future force alongside key requirements of theater posture, access, basing, and overflight to execute plans.
To make the AI Action Plan work, this process of knowledge production ought to become the beating heart of modern professional military education and senior service colleges. Faculty should do research with students, not just disseminate glittering generalities and epiphanies. Students should be pushed hard, and some will fail, to demonstrate they have the foundational expertise and can apply it to the study of major campaigns.
Most importantly, analytical processes — from deliberate planning to campaign analysis and wargaming — should be based on doctrine and domain-specific datasets the military can use to refine AI models. Foundational models are generalists by design, so they need refinement. At the most basic level, this can mean asking better questions, shaping how algorithms search for answers, and setting rules that narrow their outputs. Modern warfare, however, requires more than these basics. It calls for reinforcement learning with human feedback, where the knowledge created in the classroom is used to sharpen the models that will underpin AI agents across the force.
That means the classroom becomes a data capture site. Professional military education becomes a new type of “brain of the army” where algorithms learn from top thinkers. This pivot isn’t hard or costly, but it does require structural and cultural change.
Structurally, the schoolhouses will likely become smaller but more elite. The money saved on large administrative support staff and faculty can be directed into buying AI licenses and building the type of data capture in the classroom needed to train algorithms. This will require changing the joint regulations that push schools to dilute their focus on modern war while better aligning graduate degrees and student learning outcomes.
Culturally, the faculty will need to reconceptualize their role. There are great professors in modern professional military education, and they will have a home, but it will be one in which they both educate and research alongside students and focus on war, not pet projects. It will be a world that focuses more on practicum than core courses and embraces cooperative education and connectivism as learning theories. Students will learn by doing, and in the process, create new knowledge that refines AI models and agents used in the operating forces. And it will be a world in which faculty, by definition, should be top researchers who work with student teams to solve key operational problems.
“If I Only Had a Brain”
Military education is an old art in need of new ideas and approaches that better align it with the reality of modern war. The classroom should become a battle lab where teams analyze problems, test hypotheses, and generate new solutions to operational problems. This entire process becomes central to training AI models, creating in effect a new “brain of the army” for an algorithmic era. It is a bold reform and one that will require substantial change, ranging from how faculty and students are evaluated to the role of technology in the classroom. These changes are less a function of money and more a function of deliberate choices and thinking about the opportunity costs at play in the current design and execution of professional military education. There will be fewer administrative faculty, like deans and academic support staff, and higher requirements for aligning teaching and practice in the classroom. There will need to be more deliberate mechanisms for identifying key operational problems, moving beyond lists and technical approaches to data capture and model refinement.
The best militaries historically are those capable of reflection and innovation. They don’t rest on their laurels. They create enlightened soldiers who embrace the moment. In the 21st century, that means combining the best of human judgment with AI agents to create new ways to plan and fight wars. That process should start in the schoolhouse.
Benjamin Jensen is the Frank E. Petersen chair at the School of Advanced Warfighting, Marine Corps University and the director of the Futures Lab at the Center for Strategic and International Studies. The views in this article are those of the author and not those of Marine Corps University, the Defense Department, or any part of the U.S. government.
