Agentic Warfare Is Here. Will America Be the First Mover?

It is 02:35:01Z, March 22, 2026, inside a dimly lit Forward U.S. Joint Operations Center. Screens pulse in gentle waves each a second long, reflecting streams of updated data. Unseen beneath layers of code, an advanced military logistics AI agent combs through decades of global supply chain records, hunting for subtle patterns in what looks like chaos. Abruptly, it pauses: an anomaly. Historical pricing data on iron, copper, and coal — echoes reminiscent of the lead-up to Japan’s surprise assault on Port Arthur in 1904 — begin aligning with current market movements.

The AI agent triggers an alert, activating intelligence and planning agents that begin to work in parallel. The intelligence AI is cautious, finding evidence insufficiently conclusive. Yet the planning AI senses urgency. It swiftly activates dormant reserve computational agents, mobilizing them to run accelerated scenario testing with real-time data streams. At exactly 02:38:05Z, fresh correlations solidify — a ghostly premonition given form by intricate digital cross-checks. The system now hungers for further open-source logistics data to reinforce its predictions.

Moments later, the planning agent elevates a priority alert to the human war planners, urging immediate updates to crisis response strategies and enhanced protection for critical supply lines. Simultaneously, it discreetly signals strategic-level AI assets, prompting them to initiate national-level countermeasures — digital smokescreens designed to safeguard operational secrecy and strategic advantage.

In mere minutes, this sophisticated dance of machine intuition, historical insight, and decisive anticipation has shaped a crucial, actionable insight, silently demonstrating a fraction of the relentless autonomous effort orchestrated by AI to fortify America’s readiness.

The novelty of having large language models summarize documents and generate white papers is about to give way to AI agents that support deeper analysis and planning across joint warfighting functions. The military stands on the cusp of the next major revolution, defined by transforming computational power and algorithms into initiative and tempo. Mission command will change. Planning will undergo a transformation. Smaller staffs will aggregate and act on larger volumes of information than their Napoleonic predecessors, leveraging AI to gain a strategic advantage. In essence, the AI revolution readers of War on the Rocks have been imagining for 10 years may have finally arrived.

The world is about to enter the era of agentic warfare. In this article, we outline a theory of agentic warfare and use it to define what the United States should do to maintain its military power in this era. Although the United States may have embraced the potential of AI in the aftermath of World War II, its future is up for grabs. As two of us are executives at Scale AI, a privately held AI company that does substantial business with the Defense Department, we have both a commercial as well as patriotic interest in the United States remaining preeminent. The other author is just sick of losing wars and wants to get the next one right. Failing to adapt to agentic warfare will make the fiasco of horse cavalry fighting tanks seem like sound military judgment.

We seek here to lay out how the United States can embrace this moment to drive military innovation, specifically by deploying agentic capabilities during exercises, making infrastructure investments to ensure data is ready for AI, and rethinking the military staff structure for an agentic reality.

The Revolution

Agentic warfare is here, whether we welcome it or not. The era of military planners manually gathering limited data and compiling static crisis response options on briefing slides is over. In the next few years, the defense community will see the emergence of AI agents representing military planners, logisticians, intelligence officers, and operators that harness centuries of stored experience in real-time digital collaboration, generating uniquely effective crisis solutions for human decision-makers in seconds. This is not just an incremental improvement — it is a seismic shift. First-mover advantage in leveraging this capability will not merely ensure battlefield dominance — it will be overwhelmingly decisive at every level of warfare. It could herald the dawn of a new defense paradigm, supplanting the outdated defense-industrial complex with an agile, AI-driven agentic base. The stakes could not be higher: If the United States and free nations do not seize this first-mover advantage, they will be outpaced and outmaneuvered by adversaries who may impose their authoritarian control on a global scale.

Much has been made of the advent of generative AI since the rollout of OpenAI’s GPT-2 in 2022. Silicon Valley visionaries have told of a future where AI would dramatically reduce human workload or even allow the reprioritization of humans to higher-order tasks, leaving the mundane to machines. And yet, even with the very impressive releases of new models, most professionals do not leverage them beyond the role of a college-level chatbot used to summarize tasks or write short reports. Even in defense circles, models curated for military planning and report generation on classified networks merely augment basic human tasks — the focus of military early adopters and living on the fringes of day-to-day workflows. The close of 2024, however, introduced a new AI capability because of three critical factors: scaling computing, algorithmic efficiencies, and new methods of AI training and employment. The result is not just unprecedented models that can consistently beat the best benchmarks but a new paradigm for AI: models autonomously working together, recursively reflecting and iterating ideas between models to create new ideas outside the most educated human thinking. These collaborative models with extended memories, tied to applications and workflows, result in autonomous AI agents: generative AI models that can consider, collaborate, and create content and options for human decision. Rather than rely on a few years of a single human’s experience, one polymath AI agent incorporates a millennium of human knowledge to create new ideas and then war games these ideas with several other AI agents. Each agent has access to similar knowledge but evaluates it from different perspectives and incentives, resulting in truly unique perspectives that eclipse any individual.

AI agents, acting with human oversight, could enable the joint battle networks of today that connect global sensors to call on capabilities that can deliver force and others that can delivery other effects, such as cyber tools. Just like the analysts and scouts of the past, tomorrow’s AI agents won’t just gather information — they’ll steer sensors, track enemy movements across every domain, and flag patterns that hint at hostile intent, giving humans a critical edge before the fight begins.

Embracing the Revolution

Achieving agentic warfare first-mover advantage will require not only the adoption of new technology, but the relegation of outmoded ways of warfighting. If the U.S. military takes first-mover advantage, in the coming months generative AI agents will not simply act as chatbots for planners and operators. They will operate in parallel with warfighters. These parallel operations can occur at speeds orders of magnitude faster than humans, digesting and reflecting on petabytes of information across different military functions. Machine speed will augment human judgment. AI agents focused on intelligence tasks will iterate with planning, logistics, and operator agents to digitally and dynamically provide indications and warnings alerts of adversary changes in intent to humans — changes that would normally be far below the noise level for human observation. Hostile actions that adversaries take in the “gray zone” will be more observable and identified before the realization of the fait accompli that they so often seek to realize. If the United States moves to agentic warfare, war plans will be dynamic, with agents iterating through the night with other agents and real-world data to provide recommended planning changes and notifications of new risk tradeoffs to humans in the morning. These workflows will be complete in seconds, rather than the traditional days and weeks. Planning agents will constantly reflect on adversary strengths, weaknesses, and opportunities, in view of specific war plans that could harm the interests of the United States and its allies. These agents will recursively iterate with intelligence and logistics agents, identifying new potential indications, warnings, and deterrence options all day, every day. New options will be debated between agents, with some rejected and some selected for forwarding to humans for consideration as updates to the dynamic war plan.

Risk changes will be understood instantly, reflecting live data that is beyond human capability to synthesize. Potential response options in a crisis, generated within seconds of the earliest indications and warnings, will not be drawn from static plans, but from the history on effective options to deter adversaries based on case studies and data sets. Many of these options will be well outside the traditional paradigms of defense planners. That means an agentic force will spot what the enemy plans to do before they even act — and move first, taking steps to deter or disrupt them before they can get off the starting line. This will confound expectations and instill grave doubt about the security of adversary plans, the competence of their military planners, and the likelihood of their desired outcomes.

The implementation of AI agents will also require the development of new doctrine and staff processes. Agentic warfare could provide the moment to finally rethink antiquated staff structures, which still resemble their Napoleonic forebearers. Agentic staffs could allow for scalable oversight beyond standard test and evaluation approaches. This new approach could integrate a human-on-the-loop framework as well as automated tools to monitor agentic systems as they operate across systems. In other words, planning and operating at machine speed doesn’t mean humans are inherently “off-the-loop.”

This is just at the operational level of war at combatant commands. At the tactical level, if the U.S. military adopts agentic warfare, agents deployed to the edge could autonomously monitor sensors and signals for slight variations that could indicate an adversary’s presence before conscious detection by a human. Agents could autonomously write, review, and gain human approval of code for military tasks and cyber effects. Agentic alerting could elevate humans from observers of screens to managers of decisions and effects. Agents could monitor expenditures of munitions and fuel distribution across a battlefield recommended replenishment and resupply in advance of the felt human need heard over the radio.

Battle damage assessment will be autonomously tasked and assessed by agents making recommendations to humans and dramatically shortening the time to reattack or move to the next target. Forces operating in close proximity but lacking a common knowledge of capabilities across the services will suddenly speak the same language. Tactical planners will select an effects area and agents will recommend efficient and effective joint solutions. Tactical losses and expenditures will autonomously be compiled and communicated to higher headquarters agents with recommendations for human reinforcement decisions before humans identify a gap.

If the United States is the first mover, these tactical agentic insights will flow autonomously to the strategic level. Feeding a new defense agentic base that not only sees critical capability gaps before they are a risk in crisis but also iteratively considers new options for capabilities that are outside immediate human imagination. Gaps, seams, and solutions for global defense transportation networks will emerge from agents who not only deeply understand the warfighting requirement, but who also iterate with like agents steeped with live insights into every commodity, every port, every wind pattern.

Military resupply orders will be recommended by agents days in advance of the felt human need. Autonomous deep cyber scans will reveal previously unforeseen vulnerabilities — those of both the United States and its adversaries. With agents performing at levels above the best biochemist, the best quantum physicist, the best AI researcher, novel technologies for communications, air defense, strategic attack, and far more will emerge from these agents who think in interdisciplinary wholes at the level beyond Ph.D.s. If adopted, this dizzying flywheel of advancements will fuel a new defense agentic base to quickly render irrelevant the industrial complexes of the past.

But what if the United States and its allies do not seize first-mover advantage? The opportunity cost, in this case, is existential for the West. The same agentic capabilities employed by adversaries, even before they elevate to the level of artificial general intelligence, will dramatically outperform traditional Western paradigms of 24- to 72-hour decision cycles and wartime initiative. Once adversaries learn how to employ these capabilities and move to the level of artificial general intelligence — and, eventually, superintelligence — there will be no ability to recover lost ground. With agents guiding their military effects, these adversaries will easily outthink, outmaneuver, and dominate the West — imposing their agentic-based authoritarian control on all life. The implications for the free world are terrible to ponder. Seizing first-mover advantage in agentic warfare is a top national security concern and it is not optional.

Conclusion: Meeting the Moment

Agentic warfare represents the next military revolution and one which the United States and its democratic partners and allies can ill afford to cede to authoritarian states like China. This shift demands not only the integration of advanced AI agents into military operations but also a cultural and structural adaptation within defense organizations. It is hard to imagine waging agentic warfare through legacy military organizations effectively. Training, doctrine, and procurement processes should be reimagined to support an AI-driven paradigm that emphasizes agility, collaboration, and real-time decision-making. The time to act decisively is now. Waiting risks ceding the high ground to adversaries prepared to exploit the transformative potential of AI, thereby threatening the stability and security that the free world values and strives to protect.

To make agentic warfare a reality, the United States should to field these systems and adopt an agile approach now. Theory alone won’t cut it. The U.S. military and its industry partners need practical insights drawn from experimentation in real-world conditions. At the same time, the U.S. military should build the backbone: more computing power and infrastructure, better data, and streamlined pipelines. AI agents are only as good as the information they process. That means treating data as a weapons system — organized, accessible, and combat-ready as opposed to crap on a shared drive.

This shift also demands new thinking about military staff structures. The U.S. military should move beyond humans-only planning cells and start designing teams where AI agents act as true collaborators. That means new roles, new battle rhythms, and new doctrine. Adapting to this paradigm isn’t optional. It’s how we stay ahead in a world where AI is changing the character of war.

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.

Dan Tadross leads the public sector business for Scale AI.

Matthew Strohmeyer is a retired U.S. Air Force officer who leads public sector strategy and agentic warfare development for Scale AI.

The views here are those of the authors and not those of the Marine Corps, the Department of Defense, or any part of the U.S. government.

Image: Midjourney