It’s 02:35:01Z, March 22, 2026, inside a dimly lit Ahead U.S. Joint Operations Middle. Screens pulse in mild waves every a second lengthy, reflecting streams of up to date information. Unseen beneath layers of code, a sophisticated navy logistics AI agent combs by means of many years of world provide chain data, looking for delicate patterns in what appears like chaos. Abruptly, it pauses: an anomaly. Historic pricing information on iron, copper, and coal — echoes paying homage to the lead-up to Japan’s shock assault on Port Arthur in 1904 — start aligning with present market actions.
The AI agent triggers an alert, activating intelligence and planning brokers that start to work in parallel. The intelligence AI is cautious, discovering proof insufficiently conclusive. But the planning AI senses urgency. It swiftly prompts dormant reserve computational brokers, mobilizing them to run accelerated situation testing with real-time information streams. At precisely 02:38:05Z, recent correlations solidify — a ghostly premonition given kind by intricate digital cross-checks. The system now hungers for additional open-source logistics information to bolster its predictions.
Moments later, the planning agent elevates a precedence alert to the human conflict planners, urging fast updates to disaster response methods and enhanced safety for important provide strains. Concurrently, it discreetly indicators strategic-level AI property, prompting them to provoke national-level countermeasures — digital smokescreens designed to safeguard operational secrecy and strategic benefit.
In mere minutes, this refined dance of machine instinct, historic perception, and decisive anticipation has formed an important, actionable perception, silently demonstrating a fraction of the relentless autonomous effort orchestrated by AI to fortify America’s readiness.
The novelty of getting massive language fashions summarize paperwork and generate white papers is about to present technique to AI brokers that assist deeper evaluation and planning throughout joint warfighting capabilities. The navy stands on the cusp of the following main revolution, outlined by reworking computational energy and algorithms into initiative and tempo. Mission command will change. Planning will endure a metamorphosis. Smaller staffs will mixture and act on bigger volumes of knowledge than their Napoleonic predecessors, leveraging AI to achieve a strategic benefit. In essence, the AI revolution readers of Struggle on the Rocks have been imagining for 10 years might have lastly arrived.
The world is about to enter the period of agentic warfare. On this article, we define a concept of agentic warfare and use it to outline what america ought to do to take care of its navy energy on this period. Though america might have embraced the potential of AI within the aftermath of World Struggle II, its future is up for grabs. As two of us are executives at Scale AI, a privately held AI firm that does substantial enterprise with the Protection Division, now we have each a business in addition to patriotic curiosity in america remaining preeminent. The opposite writer is simply sick of dropping wars and desires to get the following one proper. Failing to adapt to agentic warfare will make the fiasco of horse cavalry preventing tanks appear to be sound navy judgment.
We search right here to put out how america can embrace this second to drive navy innovation, particularly by deploying agentic capabilities throughout workouts, making infrastructure investments to make sure information is prepared for AI, and rethinking the navy employees construction for an agentic actuality.
The Revolution
Agentic warfare is right here, whether or not we welcome it or not. The period of navy planners manually gathering restricted information and compiling static disaster response choices on briefing slides is over. Within the subsequent few years, the protection group will see the emergence of AI brokers representing navy planners, logisticians, intelligence officers, and operators that harness centuries of saved expertise in real-time digital collaboration, producing uniquely efficient disaster options for human decision-makers in seconds. This isn’t simply an incremental enchancment — it’s a seismic shift. First-mover benefit in leveraging this functionality won’t merely guarantee battlefield dominance — it will likely be overwhelmingly decisive at each stage of warfare. It may herald the daybreak of a brand new protection paradigm, supplanting the outdated defense-industrial complicated with an agile, AI-driven agentic base. The stakes couldn’t be increased: If america and free nations don’t seize this first-mover benefit, they are going to be outpaced and outmaneuvered by adversaries who might impose their authoritarian management on a world scale.
A lot has been product of the arrival of generative AI because the rollout of OpenAI’s GPT-2 in 2022. Silicon Valley visionaries have advised of a future the place AI would dramatically scale back human workload and even enable the reprioritization of people to higher-order duties, leaving the mundane to machines. And but, even with the very spectacular releases of recent fashions, most professionals don’t leverage them past the function of a college-level chatbot used to summarize duties or write brief stories. Even in protection circles, fashions curated for navy planning and report technology on categorized networks merely increase primary human duties — the main target of navy early adopters and dwelling on the fringes of day-to-day workflows. The shut of 2024, nonetheless, launched a brand new AI functionality due to three important elements: scaling computing, algorithmic efficiencies, and new strategies of AI coaching and employment. The consequence is not only unprecedented fashions that may constantly beat the perfect benchmarks however a brand new paradigm for AI: fashions autonomously working collectively, recursively reflecting and iterating concepts between fashions to create new concepts outdoors essentially the most educated human pondering. These collaborative fashions with prolonged recollections, tied to purposes and workflows, end in autonomous AI brokers: generative AI fashions that may take into account, collaborate, and create content material and choices for human determination. Slightly than depend on a couple of years of a single human’s expertise, one polymath AI agent incorporates a millennium of human data to create new concepts after which conflict video games these concepts with a number of different AI brokers. Every agent has entry to related data however evaluates it from totally different views and incentives, leading to actually distinctive views that eclipse any particular person.
AI brokers, performing with human oversight, may allow the joint battle networks of at this time that join international sensors to name on capabilities that may ship power and others that may supply different results, akin to cyber instruments. Identical to the analysts and scouts of the previous, tomorrow’s AI brokers received’t simply collect data — they’ll steer sensors, monitor enemy actions throughout each area, and flag patterns that trace at hostile intent, giving people a important edge earlier than the battle begins.
Embracing the Revolution
Reaching agentic warfare first-mover benefit would require not solely the adoption of recent know-how, however the relegation of outmoded methods of warfighting. If the U.S. navy takes first-mover benefit, within the coming months generative AI brokers won’t merely act as chatbots for planners and operators. They are going to function in parallel with warfighters. These parallel operations can happen at speeds orders of magnitude sooner than people, digesting and reflecting on petabytes of knowledge throughout totally different navy capabilities. Machine velocity will increase human judgment. AI brokers targeted on intelligence duties will iterate with planning, logistics, and operator brokers to digitally and dynamically present indications and warnings alerts of adversary modifications in intent to people — modifications that might usually be far under the noise stage for human commentary. Hostile actions that adversaries take within the “grey zone” will probably be extra observable and recognized earlier than the belief of the fait accompli that they so usually search to understand. If america strikes to agentic warfare, conflict plans will probably be dynamic, with brokers iterating by means of the night time with different brokers and real-world information to offer really helpful planning modifications and notifications of recent threat tradeoffs to people within the morning. These workflows will probably be full in seconds, relatively than the standard days and weeks. Planning brokers will always replicate on adversary strengths, weaknesses, and alternatives, in view of particular conflict plans that would hurt the pursuits of america and its allies. These brokers will recursively iterate with intelligence and logistics brokers, figuring out new potential indications, warnings, and deterrence choices all day, every single day. New choices will probably be debated between brokers, with some rejected and a few chosen for forwarding to people for consideration as updates to the dynamic conflict plan.
Threat modifications will probably be understood immediately, reflecting stay information that’s past human functionality to synthesize. Potential response choices in a disaster, generated inside seconds of the earliest indications and warnings, won’t be drawn from static plans, however from the historical past on efficient choices to discourage adversaries based mostly on case research and information units. Many of those choices will probably be effectively outdoors the standard paradigms of protection planners. Meaning an agentic power will spot what the enemy plans to do earlier than they even act — and transfer first, taking steps to discourage or disrupt them earlier than they’ll get off the beginning line. This can confound expectations and instill grave doubt in regards to the safety of adversary plans, the competence of their navy planners, and the probability of their desired outcomes.
The implementation of AI brokers can even require the event of recent doctrine and employees processes. Agentic warfare may present the second to lastly rethink antiquated employees constructions, which nonetheless resemble their Napoleonic forebearers. Agentic staffs may enable for scalable oversight past normal take a look at and analysis approaches. This new strategy may combine a human-on-the-loop framework in addition to automated instruments to observe agentic programs as they function throughout programs. In different phrases, planning and working at machine velocity doesn’t imply people are inherently “off-the-loop.”
That is simply on the operational stage of conflict at combatant instructions. On the tactical stage, if the U.S. navy adopts agentic warfare, brokers deployed to the sting may autonomously monitor sensors and indicators for slight variations that would point out an adversary’s presence earlier than aware detection by a human. Brokers may autonomously write, assessment, and achieve human approval of code for navy duties and cyber results. Agentic alerting may elevate people from observers of screens to managers of selections and results. Brokers may monitor expenditures of munitions and gasoline distribution throughout a battlefield really helpful replenishment and resupply prematurely of the felt human want heard over the radio.
Battle injury evaluation will probably be autonomously tasked and assessed by brokers making suggestions to people and dramatically shortening the time to reattack or transfer to the following goal. Forces working in shut proximity however missing a typical data of capabilities throughout the providers will all of the sudden communicate the identical language. Tactical planners will choose an results space and brokers will suggest environment friendly and efficient joint options. Tactical losses and expenditures will autonomously be compiled and communicated to increased headquarters brokers with suggestions for human reinforcement selections earlier than people establish a niche.
If america is the primary mover, these tactical agentic insights will stream autonomously to the strategic stage. Feeding a brand new protection agentic base that not solely sees important functionality gaps earlier than they’re a threat in disaster but in addition iteratively considers new choices for capabilities which might be outdoors fast human creativeness. Gaps, seams, and options for international protection transportation networks will emerge from brokers who not solely deeply perceive the warfighting requirement, however who additionally iterate with like brokers steeped with stay insights into each commodity, each port, each wind sample.
Army resupply orders will probably be really helpful by brokers days prematurely of the felt human want. Autonomous deep cyber scans will reveal beforehand unexpected vulnerabilities — these of each america and its adversaries. With brokers acting at ranges above the perfect biochemist, the perfect quantum physicist, the perfect AI researcher, novel applied sciences for communications, air protection, strategic assault, and way more will emerge from these brokers who assume in interdisciplinary wholes on the stage past Ph.D.s. If adopted, this dizzying flywheel of developments will gasoline a brand new protection agentic base to shortly render irrelevant the commercial complexes of the previous.
However what if america and its allies don’t seize first-mover benefit? The chance price, on this case, is existential for the West. The identical agentic capabilities employed by adversaries, even earlier than they elevate to the extent of synthetic normal intelligence, will dramatically outperform conventional Western paradigms of 24- to 72-hour determination cycles and wartime initiative. As soon as adversaries learn to make use of these capabilities and transfer to the extent of synthetic normal intelligence — and, finally, superintelligence — there will probably be no potential to get well misplaced floor. With brokers guiding their navy results, these adversaries will simply outthink, outmaneuver, and dominate the West — imposing their agentic-based authoritarian management on all life. The implications for the free world are horrible to ponder. Seizing first-mover benefit in agentic warfare is a prime nationwide safety concern and it’s not optionally available.
Conclusion: Assembly the Second
Agentic warfare represents the following navy revolution and one which america and its democratic companions and allies can ailing afford to cede to authoritarian states like China. This shift calls for not solely the combination of superior AI brokers into navy operations but in addition a cultural and structural adaptation inside protection organizations. It’s laborious to think about waging agentic warfare by means of legacy navy organizations successfully. Coaching, doctrine, and procurement processes ought to be reimagined to assist an AI-driven paradigm that emphasizes agility, collaboration, and real-time decision-making. The time to behave decisively is now. Ready dangers ceding the excessive floor to adversaries ready to take advantage of the transformative potential of AI, thereby threatening the steadiness and safety that the free world values and strives to guard.
To make agentic warfare a actuality, america ought to to subject these programs and undertake an agile strategy now. Idea alone received’t minimize it. The U.S. navy and its business companions want sensible insights drawn from experimentation in real-world situations. On the similar time, the U.S. navy ought to construct the spine: extra computing energy and infrastructure, higher information, and streamlined pipelines. AI brokers are solely nearly as good as the data they course of. Meaning treating information as a weapons system — organized, accessible, and combat-ready versus crap on a shared drive.
This shift additionally calls for new desirous about navy employees constructions. The U.S. navy ought to transfer past humans-only planning cells and begin designing groups the place AI brokers act as true collaborators. Meaning new roles, new battle rhythms, and new doctrine. Adapting to this paradigm isn’t optionally available. It’s how we keep forward in a world the place AI is altering the character of conflict.
Benjamin Jensen is the Frank E. Petersen chair on the Faculty of Superior Warfighting, Marine Corps College and the director of the Futures Lab on the Middle for Strategic and Worldwide Research.
Dan Tadross leads the general public sector enterprise for Scale AI.
Matthew Strohmeyer is a retired U.S. Air Power officer who leads public sector technique and agentic warfare improvement for Scale AI.
The views listed here are these of the authors and never these of the Marine Corps, the Division of Protection, or any a part of the U.S. authorities.
Picture: Midjourney
