Decision Dominance in the Age of Agentic AI
Introduction
Imagine a commander on a battlefield where every second is a razor’s edge between victory and catastrophe. One delayed decision, one misstep, and lives are lost, missions collapse, wars shift. This is decision dominance, the power to understand, decide, and act faster than the enemy. In today’s conflicts, where data moves faster than bullets, decision dominance is survival. Nations like China and Russia are racing to master it, pouring billions into AI and real-time networks to outpace the United States.
The Department of Defense has recognized this reality. In its 2023 Data, Analytics, and Artificial Intelligence Adoption Strategy, it emphasizes that AI integration is essential to “accelerate the speed of commanders‘ decisions and improve the quality and accuracy of those decisions.” Reflecting this commitment, the DoD has allocated $1.8 billion for AI programs in fiscal year 2025, underscoring the strategic priority placed on AI-enabled decision-making. This substantial investment signals a shift from traditional decision support systems to Agentic AI, systems that do not merely provide information but actively sense, reason, and act within command-and-control workflows.
Unlike static dashboards or predictive models, these systems continuously interpret live operational data, coordinate across multiple domains, and generate courses of action in real-time. This distinction matters: many tools branded as “AI” today are little more than sophisticated retrieval engines or generative models delivering summaries, which is far from truly autonomous, goal-oriented agents. They are the digital equivalent of early mechanization in World War I, when armies bolted machine guns onto horse-drawn wagons or strapped armor to trucks, calling themselves “armor,” incremental steps that fell short of true maneuver warfare. By contrast, Agentic AI represents a blitzkrieg leap: an integrated, adaptive system that fuses sensors, information, warfighting functions, operations with assessments as a feedback loop. This creates a dynamic decision support system that moves at the speed of conflict.
As the Joint All-Domain Command and Control (JADC2) Strategy emphasizes, achieving “information advantage at the speed of relevance” requires precisely this kind of AI-enabled agility,. History shows that militaries often diverge in how they respond to disruptive technology. Some adapt doctrine to exploit it fully, while others remain anchored in past methods, solving yesterday’s problems while their adversaries prepare for tomorrow’s.
In 1940, the French Army was the world’s finest, armed to the teeth, dug into fortified lines. Yet in six weeks, the German Wehrmacht crushed them. Why? The French clung to static defenses and slow couriers, tethered to old tactics. The Germans, by contrast, rewrote their doctrine in Truppenführung and built a force designed to exploit the cutting-edge technology of the day: FM radio. This simple but revolutionary tool allowed dispersed tanks, infantry, and aircraft to communicate and synchronize at a tempo the French could not match. For the first time, commanders could practice true operational art, coordinating maneuver across multiple domains in real time. They turned chaos into victory, achieving decision dominance that left the French reeling.
Today, the choice is similar. Armies can embrace live data and Agentic AI, reshaping doctrine and practice to fight the wars of the future, or remain anchored in legacy methods, solving the problems of the past. Just as the FM radio unlocked operational art in 1940, Agentic AI has the potential to unlock decision-centric warfare today. The question is whether doctrine and habits will adapt to seize it.
Obstacles and Enduring Challenges
Yet this leap is not without obstacles both culturally and technically. Integrating agentic systems into staff processes requires overcoming organizational resistance and entrenched habits. History shows that militaries often struggle to abandon legacy methods, and the introduction of AI is no exception. Much has been written about how AI could reshape the Napoleonic staff model itself, raising the possibility that decision-making processes such as the Military Decision-Making Process (MDMP) may evolve into something fundamentally different. The greater question, however, is one of trust: will commanders and staffs accept AI-generated assumptions, courses of action, and recommendations when lives are at stake? Even the most sophisticated system cannot confer advantages if human leaders hesitate to use it.
The greater question, however, is one of trust: will commanders and staffs accept AI-generated assumptions, courses of action, and recommendations when lives are at stake?
A further complication is the crowded landscape of so-called “AI” tools. Many systems marketed as artificial intelligence are, in practice, little more than retrieval-augmented generation models or static dashboards wrapped in the language of autonomy. These imitators create noise in the operational environment, leading commanders and staffs to question whether any AI system can truly deliver on its promises. When inflated claims fail in practice, confidence in the broader concept of AI is undermined, making leaders more hesitant to adopt agentic systems even when they offer genuine capability. The danger is not just technological confusion but institutional skepticism. Clear definitions, rigorous evaluation, and demonstrated performance are therefore essential if Agentic AI is to be distinguished from imitators and trusted as a partner in planning and operations.
Beyond culture, technical integration poses its own challenges. Connecting agentic systems across services, echelons, and security domains introduces friction, as legacy platforms, classification barriers, and competing standards complicate the flow of information. At the same time, reliance on AI introduces its own form of friction. Adversaries will look to exploit this new mode of decision-making. They will seek to deceive, inject false data, or corrupt the very sources AI depends on, which becomes a form of digital fog of war. The challenge, then, is not simply building intelligent systems, but ensuring they are resilient against manipulation.
The hardest challenge, however, is not building the system but convincing leaders to act on its outputs. That hesitation is not new. For centuries, commanders and staffs have wrestled with how to see clearly in the fog of war and decide with confidence despite uncertainty. Carl von Clausewitz captured this enduring problem in his 1832 masterpiece On War, describing coup d’œil—the commander’s “inner eye” that discerns the decisive elements of a situation at a glance, even amid confusion and friction. He warned that without this faculty, plans collapse under pressure, because clarity is the foundation of all decisive action.
These barriers do not diminish the promise of Agentic AI. In fact, they highlight that successful adoption must be accompanied by cultural change, secure integration, and doctrine that frames AI as a trusted partner rather than a disruptive outsider. Overcoming these hurdles will require not only technical solutions but also training, experimentation, and organizational reform. These are all steps that ensure commanders and staffs can rely on AI outputs with confidence rather than suspicion. The payoff is significant: a force that can exploit the speed and adaptability of agentic systems without sacrificing the trust and judgment that remain central to command.
In this sense, the problem facing commanders and staffs today is less about technology than about clarity and judgment. Agentic AI can be understood as an extension of the coup d’œil: a tool that senses, interprets, and presents live operational data in a way that accelerates comprehension without supplanting human intuition. Where Clausewitz emphasized the commander’s cultivated ability to grasp what matters most, AI can expand that faculty across larger, faster, and more complex streams of information. This framing aligns directly with the Army’s doctrinal steps of command—Understand, Visualize, (Describe), Direct, (Lead), Assess. Each element reflects a modern embodiment of coup d’œil. Trust, then, comes not from replacing intuition with a machine, but from fusing human intuition with AI-enabled perception to achieve a sharper, more resilient form of coup d’œil at every stage of the decision cycle.
Through this historical lens, the subsequent sections illustrate how modern tools, coupled with classical principles of command, can produce decision dominance in practice. The following stories are drawn from firsthand command experience, showing how clarity, visualization, decisive action, and continuous assessment can be enhanced through technology to outpace adversaries in complex operational environments.
Understand — Clarity Before Action
Clausewitz described coup d’œil as the ability to see clearly through uncertainty, a skill that remains the first step in achieving understanding in modern operations. In January 2022, during planning for a looming crisis in Ukraine, the XVIII Airborne Corps faced how fragile that clarity can be. Tasked with planning a Non-combatant Evacuation Operation as Russia’s invasion of Ukraine loomed, the team needed an airfield in Poland to serve as a Temporary Safe Haven for evacuees and incoming C-17s. Given Poland’s long-standing status as a NATO ally, it was expected that critical details—runway specifications, fuel capacity, and infrastructure, would be readily available from U.S. European Command. They were not.
Instead, the planning team found itself circling potential sites on a paper map and then spending the better part of a day digging through scattered sources to research each location, trying to determine which were feasible. Even after that effort, there was no substitute for sending leaders forward to physically recon each site, an approach that consumed weeks when days mattered. The gap in knowledge and understanding nearly derailed the operation before it began. It was a vivid demonstration of Clausewitz’s insight: commanders cannot rely on secondhand reports or assumptions, but must cultivate a grasp of the environment, or coup d’œil, to discern decisive elements at a glance.
The broader lesson is just as important: assuming higher headquarters will provide everything needed is a mistake. There is no shortage of information today, but its volume is overwhelming, and identifying relevant details in time is the true challenge. Leaders usually know the decisions they must make and the information required, but finding and validating that information through manual processes takes precious time. In war, time is a commodity that cannot be wasted.
What once demanded laborious reconnaissance could be compressed into seconds, enabling understanding at the speed required for modern conflict.
Agentic AI offers transformational potential. By fusing live radar feeds, open-source data, social media streams, logistics databases, and other sources, an agentic system could have delivered instant clarity on Poland’s airfield runway conditions, throughput capacity, and infrastructure status, without weeks of delay. What once demanded laborious reconnaissance could be compressed into seconds, enabling understanding at the speed required for modern conflict. The ability to understand faster than the enemy is the first step toward decision dominance.
Visualize — Seeing the End State Before the First Move
Understanding the environment is only the first step. Commanders must also visualize the fight, cutting through complexity to imagine the end state before the first move. This step often requires time and reflection, but in modern war, time is scarce. Agentic AI offers an advantage: by processing vast amounts of data, surfacing recommendations, and testing assumptions, visualization can be achieved faster and with greater clarity, accelerating planning and achieving decision dominance.
In September 1918, Colonel George C. Marshall faced this challenge during the Meuse-Argonne Offensive, a make-or-break campaign to end World War I. As the American Expeditionary Forces (AEF) Plans Officer (G5), the task was staggering: move 600,000 troops, 3,000 artillery pieces, 90,000 horses, and a million tons of supplies 50 miles in two weeks, without alerting the Germans. Overwhelmed, he stepped away. Sitting by a French canal, he watched an elderly fisherman cast his line. For half an hour, silence. Then, clarity: the network of roads, the cadence of night marches, and coordination with French allies, a logistical masterpiece began to taking shape. Back at his desk, he turned vision into plan, moving the American Expeditionary Forces undetected and setting the stage for victory. Historian Edward G. Lengel calls it a turning point.
Marshall’s canal-side vision demonstrates how deliberate focus under pressure can unlock clarity, but also highlights the time required to reach insight. Today, agentic systems can compress that process, turning hours of contemplation into near-real-time visualization. By surfacing options, stress-testing assumptions, and helping commanders and staffs see the end state sooner, Agentic AI accelerates the planning cycle, giving leaders the ability to visualize faster than the enemy and seize decision dominance.
Direct — From Insight to Action
Directing is where insight becomes action. It is about making decisions decisively, adapting under pressure, and turning possibilities into results. Visualization helps identify options, but directing turns understanding into outcomes.
In March 2023, during non-combatant operations in Sudan, Khartoum’s only viable airport was blocked, and hundreds of Americans needed evacuation quickly. For days, every conventional option was unattainable and the outcome seemed impossible. Then a memory sparked: a friend laughing about Planes, Trains, and Automobiles, Steve Martin yelling, “Those aren’t pillows!” after an awkward moment with John Candy. That line triggered a realization: planes were out, but trains, buses, or boats might work. The contracting team was directed to explore these alternatives: “Check train tickets, charter buses, find coastal routes.” A nonstop bus from Khartoum to a port was secured, followed by boats to Saudi Arabia. It was not pretty, but it saved hundreds of lives.
The lesson: directing is about seizing opportunity and turning inspiration into action, even when the situation seems impossible. Agentic AI could transform this process. Instead of relying on human memory or improvisation alone, AI could rapidly query live transport data, generate multiple viable routes, and present actionable options in seconds. What once required intuition and persistence can now be achieved faster, with greater clarity and confidence, turning sparks of insight into coordinated, decisive action.
Assess — Continuous Evaluation
Directing gets plans moving, but execution without follow-up is incomplete. Assessment ensures plans deliver. It is not a one-off; it is a relentless cycle of evaluating actions, spotting gaps, and improving, even when no one is complaining.
This lesson was reinforced in 2017 during operations against ISIS in Iraq and Syria. A High Mobility Artillery Rocket System (HIMARS) battalion supporting those operations had been operating for weeks. The strikes it was delivering seemed successful. But based on the commander’s observations after a battlefield circulation, challenged the assumptions that all strikes were successful. The unit began to preform continuous assessments through the Battle Damage Assessment (BDA) process for every strike. These assessments uncovered a problem: rockets from one area were off by five to ten meters. Close enough to achieve effects, but in a city, five to ten meters could be catastrophic. Investigation revealed a flaw in the Advanced Field Artillery Tactical Data System (AFATDS). The AFATDS defaults to Height Above Ellipsoid for target elevation, instead of what the HIMARS launcher uses: Mean Sea Level. This difference led to an error in elevation calculations that ultimately offset the rockets by five to ten meters. No one had complained, but active assessments identified the error, enabling correction and restoring precision. Assessment was not a formality; it was a safeguard against unseen risk.
Today, agentic AI could perform this kind of assessment continuously, analyzing live BDA feeds and operational data, highlighting anomalies, and delivering actionable insights in real time. Instead of waiting for the next battle rhythm event or decision board, commanders could act immediately to correct course, reduce risk, and maintain decision dominance. Assessment is continuous, adaptive, and far faster than humans alone.
Conclusion — Agentic AI and the Future of Decision Dominance
Decision dominance in the future will depend on executing the cycle of understanding, visualizing, directing, and assessing at a tempo faster than the enemy. The vignettes from Ukraine, the Meuse-Argonne, Sudan, and the fight against ISIS all underscored how fragile and decisive each of these steps can be: clarity before action, vision under pressure, decisive direction, and relentless assessment. What changes today is not the importance of the cycle, but the speed at which it must be executed. Adversaries are already harnessing AI and digital networks to accelerate their own decision processes, compressing timelines and raising the stakes. To maintain an advantage, commanders and staffs must integrate agentic AI as a partner that enables each step of the cycle to happen faster, with greater clarity, and at a scale no staff alone can match.
Agentic AI can continuously integrate live data, analyze outcomes in real time, and present actionable options. These systems can accelerate understanding, compress visualization, enhance the speed and precision of guidance given, and make assessment persistent rather than periodic. Decision dominance is no longer merely the art of the commander’s intuition; it can become a sustained operational capability when powered by autonomous, adaptive intelligence. Realizing this potential, however, requires more than just adding AI to a workflow.
Exia Labs products like Blue (pictured) and Recon leverage Agentic AI to understand complex operational environments and support the warfighter by creating an understanding of the operational environment, and generating courses of action with actionable recommendations. This next-generation planning technology accelerates understanding and decision-making.
The solution lies in addressing cultural, technical, and institutional challenges together. For the cultural barriers of trust and adoption, leaders must normalize AI through training, education, and wargaming that treat agentic systems as staff teammates rather than mysterious black boxes. For the technical barriers of integration, systems must be tested, red-teamed, and designed for interoperability across domains and classification levels, with resilience against deception and manipulation built in from the start. And for the challenge of imitators, the Department must establish clear standards, rigorous evaluation processes, and transparent demonstrations that separate genuine capability from marketing hype before credibility is lost.
In practice, this means doctrinal adaptation and even reorganization, embedding agentic AI into staffs as an accountable member of the decision-making team, much as past generations integrated Field Artillery officers, Intelligence Analysts, or Logisticians. By doing so, the institution can overcome mistrust, enforce discipline in its use, and ensure commanders and staffs see it not as a threat but as a force multiplier. The payoff is a decision process that is faster, sharper, and more adaptive, enabling decision dominance without sacrificing the human judgment at its core.
A handful of companies are already pushing the frontier of true agentic AI, building adaptive agents, leveraging advances in gaming technology to enhance their functions, and moving far beyond the retrieval-augmented generation systems too often passed off as “AI.” These efforts point to what the future of operational planning and execution could look like if innovation is given room to grow. For the Department, the task is to identify and engage with these pioneers rather than default to organizations with flashy marketing, inflated statistics, or venture capital gloss. The real breakthroughs will come from those designing systems for the wars of tomorrow, not repackaging tools for the problems of the past.
The lesson is clear: the future of warfare will demand the deliberate integration of Agentic AI into every stage of planning and execution. Nations that embrace this technology will gain the ability to act faster, anticipate outcomes more accurately, and seize initiative in ways traditional methods cannot match. Agentic AI is not a distant possibility, it is already here. Those who act now, adapting culturally and organizationally, will master it first and secure decision dominance, while those who hesitate will fall behind in the wars of tomorrow.
