The artificial intelligence revolution has arrived in military education, and resistance is futile. AI now stands poised to fundamentally transform how we educate military professionals — whether educators are ready or not.
Moreover, those leaders of professional military education institutions scrambling to establish AI policies are fighting a losing battle. The technology is evolving faster than any bureaucracy can respond, leaving many schools with contradictory approaches that range from outright bans to ad hoc permissions. Thus, students are reporting receiving mixed messages, with some being warned they could lose academic credit for using AI tools.
This approach is both unrealistic and counterproductive. AI tools have become too powerful and accessible to ignore. Expecting students to avoid them is like asking them to abandon search engines. Moreover, students recognize that AI proficiency will be essential in their post-graduation careers.
The evidence is clear: 88 percent of college students use AI multiple times weekly and even high school students regularly employ these tools despite restrictions. Military education students are no exception. My own belief, after querying many dozens of students, is that well over half of them are using AI at some rudimentary level. The uncomfortable truth for professors is simple — it’s impossible to monitor or control AI usage effectively.
Critics argue that AI dependency will undermine writing skills. One professional military professor told me that he dedicated up to 20 hours weekly to writing instruction. This represents hundreds of academic hours annually diverted from specialized military education to remedial writing — hours wasted when writing centers already exist for this purpose. More importantly, this concern is increasingly irrelevant as AI tools will handle most professional writing throughout these students’ careers.
The fear that AI will replace critical thinking is just as misplaced. When properly integrated, AI tools can dramatically enhance critical thinking by providing sophisticated data analysis, visualizing complex concepts, generating diverse perspectives, challenging assumptions, facilitating deeper engagement, and identifying biases. Moreover, the AIs can accomplish these tasks at incredible speeds and, even more crucially, they are always available for any student wishing to undertake a moment of personalized learning.
Military education, therefore, faces a fundamental choice: embrace AI’s transformative potential or futilely resist change while students adopt these tools anyway. The institutions that thrive will be those that teach students to leverage AI effectively — using it to augment rather than replace human judgment.
Instead of asking whether AI belongs in military education, we should be asking how to harness its capabilities to create more capable, adaptable military leaders. The future of military education doesn’t lie in preserving traditional methods but in pioneering new approaches that combine human wisdom with AI to prepare the next generation for increasingly complex security challenges.
The AI tsunami is already making landfall. The question is whether professional military education will ride the wave or be swept away by it.
Interesting Use Cases This Year
This academic year has been a year of experimentation. It all started when a former Marine Corps War College student sent me a text asking if ChatGPT-4 was something to pay attention to. I played with it for an hour that evening and thought it was interesting, but it was probably not a game-changer for anything I did. Still, I came back to play with it a few days later, and after 10 hours I was convinced that the academic world and my small part in it was about to be turned upside down.
My crucial problem was that I had no idea how to exploit AI capabilities. What was painfully evident from the start, however, was that “today” is the worst the AI systems will ever be and that their capabilities are improving at a breathtaking speed. Throughout the year, as the AIs’ capabilities improved, so did the number of use cases I (and my students) discovered. The downside was that long-term planning became impossible, as the speed at which the AIs were advancing made a hash of all my early AI-employment decisions. Thus, I decided to open the spigots and let students use AI unfettered in the classes I controlled while I continued experimenting with different AI models and their expanding capabilities. Along the way, I sent out regular “Artificial Intelligence Notes” to most of the University. These were warmly received by most recipients, but certainly not everyone. Fortunately, the war college director and the university’s commanding general were early believers and strong advocates for expanding AI use throughout the university.
At this point, it may be edifying to see a few examples of how AI has been used at the Marine Corps War College this academic year. Here is one example that caught me by surprise. For the school’s semester-end oral comprehensives, one student used AI tools to analyze my online writings, predict likely exam questions, and generate concise answers. I was astonished to discover that all four questions I asked appeared on his AI-produced list. Another group streamlined their class presentation by feeding their research to ChatGPT, generating an essay, transforming it into a 20-slide presentation with Gamma, and then returning to ChatGPT for slide-specific talking points — all in under half an hour.
A different student team crafted a 6,000-word essay on the 1777 Philadelphia Campaign almost entirely through AI: Perplexity provided extensive research materials, including online primary sources; CustomGPT generated answers and structured outlines; and OpenAI’s Deep Research wrote detailed sections, which ChatGPT seamlessly integrated into a coherent final paper complete with AI-generated artwork. The team even created a now-obsolete 30-point guide for peers interested in this approach. If I tried to reproduce all of their work today, it would take me about four hours, and almost all of that time would be spent collecting the research material to feed into the AI.
This year, with 10 new courses, I significantly reduced my preparation time by employing multiple AI language models (ChatGPT, Claude, Gemini) to produce summaries, class notes, and synthesized talking points from assigned readings, cutting prep from days to about an hour per class. Recently, I distributed the Intel community’s 2025 Threat Assessment to my class. Within 30 minutes, students responded with AI-generated analyses comparing it to the 2024 assessment, aligning it with the current National Defense Strategy, and critiquing it based on curated essays shaping modern wargaming.
To enhance strategic wargaming exercises, I developed a CustomGPT modeled after China’s leader Xi Jinping, incorporating his public speeches and biographical data, enabling digital role-play of Xi’s decisions during wargames. Additional CustomGPTs we developed for professional military education simulate expertise in competition between the United States and China in the Pacific and historical great-power competition.
A New Paradigm: Doubling Down on What Works
The success of this academic year’s AI experiments encouraged me to go all-in on AI in the future. That, of course, means that much of what I currently do has to change.
What’s Going Away
The first removal is an easy decision: Writing assessments are banished forever. I will still have students write for various projects, but I expect them to use AI heavily. Then, by having them turn in a list of their prompts, I can actually track their critical thinking as they proceed through the assigned task.
If I can’t assess writing, what’s left?
To a degree, I am still working through this problem, but what I am sure of is that in AI-world, assessments will have to be made almost entirely through hugely increasing contact with students. That brings us to our next significant change; for most of my classes, I am also doing away with large seminars in favor of something approaching an Oxford tutorial. Each seminar will be broken in half so that I am teaching two groups of 7–8 people, instead of 15–16. In such an intimate setting, there will be no place for students to hide, making it easy to assess who has mastered the material and who is faking it.
Also, like an Oxford tutorial, I am no longer in the business of imparting knowledge. Going forward, I expect students to teach themselves – true active learning – before they enter the classroom. Never again will I have to pretend to be conducting a Socratic seminar, where I do almost all of the talking and everyone makes believe that answering a few student questions is an intellectual give-and-take.
I know that professional military education professors give a lot of lip service to employing the Socratic method but, within professional military education, I almost never see it done. And that is for a very good reason: It is nearly impossible to use the Socratic method when the student’s depth of knowledge on any particular topic is almost always limited to what they most likely read the night before.
Lectures have always been the easy button for professors and students. They don’t demand much from the professor, as once the notes are made and the PowerPoint is built, most lectures can be put in the rinse and repeat cycle for years. They are also good for the ego, for what professor does not want to show off how much we know to a trapped audience. Unfortunately, most — close to all — lectures have the mental shelf life of a firefly, presented in an hour and mostly forgotten in the next hour.
For decades, we have known the most effective learning methods are active learning approaches (wargames, staff rides, well-prepared case studies) coupled with self-directed learning, where the students teach themselves at their own pace. I have been employing active learning techniques to the greatest extent possible for well over a decade. Now, thanks to AI tools, I can also employ self-directed learning throughout the course. Thus, I will transition to become a “learning coach” while the students teach themselves and then have an opportunity to impress me with their knowledge.
What’s Coming On
To accomplish this, I am using AI tools to create student learning modules for each segment of my courses. Let’s take an example. I plan on teaching multiple classes on the Napoleonic era in the next academic year. To help the students teach themselves, there will be a “Napoleon Folder” they can access on the school’s shared drive. Inside this folder will be several subfolders.
The first of these folders contains the course card and class notes. The course card sets out the learning objectives, while the class notes spell out how the students will meet expectations. The class notes also lay out what is available to help them. This includes links to any informative videos, maps, and other helpful material. Most crucially, the class notes include a link to a CustomGPT on the Napoleonic Era (give it a try). I am building a separate CustomGPT for every topic I teach. All of them are filled with carefully curated published material and have an embedded prompt to help answer student questions. Not only does the CustomGPT know almost everything there is to know about the Napoleonic era, but it is also available to students 24/7, every day of the week. Note, I am using OpenAI’s GPT-4o for this, but most major large language models have similar tools (e.g., Google Gemini’s Gems or Anthropic’s Projects inside of its Claude models). For some courses, such as World War II, there will be multiple CustomGPTs, each focused on a different part of the whole (e.g., one will have an operational focus, while others will focus on economics/mobilization, wartime politics, or diplomacy and war termination).
A separate folder contains all of the student readings, which includes every reading required along with a separate folder for supplemental (optional) readings. I have been increasingly moving away from giving students essays or individual chapters to read in favor of having them read entire books. To make sure they have the time to do this, I am teaching fewer topics, but for those crucial topics that remain I am adding much more depth, which takes multiple class periods.
Still, there is only so much time in a day, so to help them get through the assigned books and to focus on the key topics, I have divided each book into 5 to 7 sections and created an AI-produced study guide for each section. For those few times when I still use shorter readings (essays, chapters), they will also each have an AI-produced study guide.
In the final folder is a series of AI-produced audio supplements. I am using NotebookLM (a Google product) to take each section of the required readings and create audio podcasts for each. This allows students to listen to two AI-podcasters discuss the big themes within the readings. These podcasts are not yet perfect, but they almost always hit on all the key themes I would have addressed in my old-style seminars. I also plan to use ElevenLabs to create a number of other audio projects during the course of the year. The intent is to allow the students to learn during periods when they have previously been unable to study (commuting, exercising, etc.).
These class folders will be available to students at the start of the academic year, allowing them to set their own pace for learning the material. Thus, when they arrive in class, I expect them all to have achieved a level of expertise that allows for much deeper discussions than have been possible before.
I am still pondering my exact role and the overall format for class sessions. But I think I will start seminars with a 10- to 15-minute talk to set the key themes and ideas that I would like to hear discussed and then elicit ideas on additional themes that caught the interest of students. From that point on, my role will be to facilitate the discussion with a light touch, to move the topics along, correct crucial points of fact, and insert some probing questions when discussion slows. My overriding goal is to ensure that most of these discussions take place between the students with as little reference to myself as possible.
Not all classes will follow the Oxford tutorial format, as I plan to enhance my small group classes with seminars designed for the students to work through specific case studies. These will be team events and follow the Harvard model to the greatest extent possible. There will also be periodic staff rides and wargaming will, of course, continue. Finally, I am not so naïve as to think I can completely do away with lecture-based classes. But they will be few and far between.
AI and Professors
I worry most professors will continue to shun the use of AI to the greatest extent possible. I sincerely believe, however, that those who take this route are doing themselves and their students a disservice. Besides, the battle to restrict AI in professional military education classrooms is already lost. Still, I do not believe that AI can replace Title 10 professors (yet), and may never be able to do so. However, I am certain that AI will soon make the bottom half of all professors superfluous and that any professor who does not integrate AI will be replaced by a professor employing AI. Before another year is out, professor-AI teams will be the elite of every university, professional military education included.
There is little in the world of academia that the AI cannot do. For instance, I gave an AI a few sentences of instructions to teach it to grade papers the same way I would (a personal rubric). I then ran 30 student papers from a few years back through an AI tool. Amazingly, the AI went 30-for-30 and gave every paper the same grade I did. I then asked the AI to give me five strengths and weaknesses of the papers, focusing on the key arguments made by the students. The result was sobering: The AI, in about half the cases, did better than I had and at least as well in all of the rest. More unsettling, it took about a minute to do each task.
In addition to some of the AI uses outlined above, AI tools can also reduce the time professors require for many other crucial tasks by orders of magnitude. For instance, AI tools can put together entire syllabi, along with recommended readings and assignment ideas, in moments. I now use my AI-buddies to prepare detailed lecture notes based on my course material and to help create case studies to support classes. Another Marine Corps University professor is using AI tools to help design one of the more complex wargames our students use.
Regarding my research, writing, and publication, AI tools have upended everything I used to do and increased my productivity by a seriously high multiple. As a sidenote, this essay may be the last essay or book I write without AI help (except to run Grammarly on the finished draft).
Last week, I started researching a new book. As a first step, I broke down the central themes and then listed all of the key characters that will appear in the work. I then had OpenAI’s Deep Research do a separate paper on each theme and person. The result was a slew of essays of 4,000–10,000 words, all well-sourced. I then asked the AIs to recommend the best source material for each theme and person. About half the sources were in one of four languages, none of which I can read. Fortunately, the AIs can read and translate them in minutes.
Still, the AIs are not perfect. For instance, they cannot look behind paywalls, so they are unable to access databases that contain academic research products like JSTOR. And you always have to check their work. It is best, therefore, to think of them as really smart Ph.D.-level interns or research assistants who have to be checked up on. Moreover, AIs do not negate the need for archival research for academic books.
What the AI did do, though, was — in a single weekend — to provide me with foundational research that previously would have taken me months to assemble. The AIs even summarized each work for me and gave me a ton of ideas on how to approach different book themes from unique perspectives. Additionally, if you are doing archival research and run across notes scribbled in illegible handwriting, an AI can decipher it instantly and turn hard-to-read fonts or even medieval script into a Word document in only moments.
I plan to use AI tools for data analysis, data visualization, citation fixing, and summarizing research material. I also plan to let an AI review my draft. Although any large language model can undertake this task, I have created another CustomGPT that will look at written works in the role of a peer reviewer for an academic press. This CustomGPT looks for things such as novelty of ideas, innovative approaches, data credibility, clarity, weaknesses and gaps in the work’s arguments, up-to-date references, and much more. It also gives the paper a grade between one and six, similar to an actual peer reviewer. Finally, it will offer concrete recommendations to improve the work. To my way of thinking, anyone who submits a paper or book to an academic publisher without using this or a similar tool first is being foolish.
Conclusion
AI represents a transformative force in professional military education, crucial for maintaining relevance, enhancing efficiency, and radically improving the depth of learning. As AI advances exponentially, resistance or ambivalence toward its use within professional military education classrooms is not merely impractical — it is strategically negligent. AI tools already excel in critical educational functions such as data analysis, personalized learning, visualization, and real-time feedback, capabilities far surpassing traditional educational methods.
The implementation of AI empowers instructors to transition from conventional lecture-based teaching to roles as active learning coaches, thereby fostering genuinely student-centered and self-directed learning environments. AI’s integration enhances critical thinking by encouraging students to interrogate, refine, and develop insights through iterative interactions with intelligent tools. Moreover, by delegating routine tasks such as curriculum development, class preparation, and grading to AI, educators gain substantial time to dedicate toward meaningful engagement with students, enhancing the quality and impact of interactions.
In the rapidly evolving professional contexts professional military education students will enter, proficiency with AI tools will not be optional but essential. Professional military education institutions that proactively incorporate AI will produce graduates far better equipped for strategic decision-making, adaptive problem-solving, and innovative leadership. Conversely, institutions or educators resisting this shift risk rapidly becoming obsolete.
Ultimately, the critical integration of AI into professional military education promises not only to revolutionize educational methodologies but also to create adaptive leaders who are thoroughly prepared to confront the complexities of future military and strategic landscapes.
James Lacey, Ph.D., is the Horner chair of war studies at Marine Corps University. He is the author ofThe Washington War and the forthcoming book, The Year God Died.
The author used ChatGPT4 to draft portions of this article, especially the conclusion.
Image: Midjourney.
Commentary