Photo: Milan_Jovic/Getty Images
The United States has repeatedly proven its ability to execute decisive kinetic strikes. Adversaries understand this, which is why they increasingly avoid conventional confrontation. Instead, they operate in the space of irregular warfare: leveraging autonomous systems, conducting cyber and infrastructure attacks, and directing cognitive campaigns shaped by AI-driven narratives aimed at influencing populations and undermining legitimacy.
Despite attempts to catch up, the United States is behind on irregular warfare. Against the backdrop of continued hostilities in the Middle East, this analysis frames a critical challenge within irregular warfare: how to better leverage data and AI tools to win the cognitive domain.
Irregular warfare is not a new concept. The contest for hearts and minds has existed from Augustus to Petraeus; however, recent conflicts, from Afghanistan and Iraq to the ongoing tensions with Iran, highlight a persistent gap. The United States dominates conventional and tactical spaces but struggles to define, much less achieve, desired end states, including longer-lasting periods of peace between violent conflict. Despite battlefield success, the failure to align military, diplomatic, economic, and cognitive tools toward a coherent end state has repeatedly limited the ability to secure lasting outcomes.
The United States possesses the models, compute, resources, and talent to compete offensively in the information environment; however, it has not demonstrated the same willingness as Russia, China, and Iran to engage in cognitive warfare, often leaving uncontested space for adversaries deploying AI-generated content at scale—including, of course, Iran’s recent viral videos.
As conflict simmers in Iran and beyond, military and intelligence leaders should assess the cognitive ground already lost and determine what it will take to regain initiative, including the intentional use of data to power agentic AI. As with any campaign, the first step remains defining strategic objectives before attempting to compete.
The United States’ inability to win peace starts with failing to define success.
Take the conflict with Iran as an example. There are Iranian supporters of U.S. and Israeli actions. Many of them live outside of Iran, but research suggests that there is a small but vocal minority of anti-regime voices using virtual private networks, proxies, and other available tools from within the country to circumvent censors and share their views. However, these same signals show that there is a larger group of anti-regime voices who are displeased with Operation Epic Fury. They hate the regime, but they hate the United States more, a situation that suits Iranian generals just fine. Despite plenty of evidence suggesting that kinetic air power alone has no recorded examples of facilitating peace—much less regime change—these disaffected Iranians are the people who President Trump initially hoped would reinvigorate the protests of early 2026 and seize control of their government. Instead, the bombings, resilient leadership structures, shifting U.S. priorities, and the persistent information campaigns waged by Iran on its people and externally have soured the very Iranians on whom the president was initially relying.
Winning the cognitive domain requires first redefining success as sustained influence over public and diplomatic opinion, with an aim to reduce the possibility of escalation with adversaries and shift toward cooperation—or at least coexistence. Data is a major underappreciated difficulty in achieving that success.
Measuring incremental and strategic success in the cognitive domain is challenging: With intangible attributes like belief and perception, how does one definitively know if efforts present return on investment or even measurable results?
The answer begins with data. The United States possesses vast amounts of information, yet very little of it is purpose built. Most datasets are adapted from commercial markets or legacy intelligence frameworks that rely on overwhelming big data, which is expensive to decipher and often irrelevant. More importantly, these streams are often optimized for winning in a kinetic fight, not winning the cognitive domain. The result is that signals are noisy and expensive, leaving unstructured and hard-to-access environments underexploited. Consequently, planners operate with incomplete pictures of the systems they seek to influence.
The goal must be to feed better, more intentional data into analysis engines. Intentional data is not collected; it is designed. It begins with clearly defined strategic outcomes and builds collection, aggregation, and processing mechanisms aligned to a singular goal and clearly defined attributes of change. Today, we can discern human characteristics, attitudes, and various types of behaviors and beliefs as tangible metrics. Via agentic AI and ever-improving processing power, human behavior is targetable at the community—and even individual—level.
Systems must have the ability to process intentional data at scale and at the speed of relevance. Rapid advancements in agentic AI, particularly those powered by models that go beyond lean structures and static training datasets, mean that the capabilities exist today to meet the challenges outlined above. But technology without strategy compounds the issues of data and AI models. The speed, scale, and velocity of future data are only useful in service of a defined and codified strategy, one that ensures the agents have specific, intentional tasks; provide truly measurable results; and have humans in the loop at multiple points along the journey.
Reforming how and when humans make decisions within these decision loops is critical if agentic AI is to be deployed effectively and with trust in cognitive warfare. Human and agent actions must be tied to clearly defined strategies and outcomes, with mechanisms that measure progress against defined attributes of change, employing the next level of data and AI constantly pushing the bounds of capability. This requires integrating intentional data, suitable authorities, and operational design as a forethought, not an afterthought.
Winning the cognitive domain requires the United States to play the game adversaries are playing, thoroughly understanding and appreciating the unique nature of each conflict and its actors. And yet, adversaries continually outperform the United States in the cognitive domain, taking the fight to Americans and their allies at home and abroad. Realizing this deficit, the Department of Defense published an instruction on irregular warfare that better defines coercion in environments where success is more subtle and difficult to achieve. According to an official involved in issuing the instruction, an accompanying irregular warfare strategy document was drafted and coordinated informally across the department and shared with a panel of experts from defense and academic communities. This strategy should be reviewed by the Trump administration and issued immediately to give the military services and combatant commands the tools with which to operationalize it.
Though it is easier to define winning when dealing with conventional military terms like degrade, deny, disrupt, destroy, delay, and manipulate, these are insufficient given the current state of competition and the increasingly irregular nature of warfare. Part of the Trump administration’s review should include coordinating a more realistic and useful definition of winning the cognitive domain. For example:
Success should thus be seen as a constant state of competition, wherein the adversary believes it is in its best interest to cooperate and compete. Irregular warfare strategies and tactics should not be designed to close the door on adversaries; they should be designed to create positional advantage, staying ahead in a perpetual—not terminal—game of strategic competition.
The United States has spent decades optimizing for conflict without fully defining success in competition, much less in irregular warfare. The result has been tactical excellence paired with strategic ambiguity, exactly what can be seen in Iran. Reframing winning as sustained advantage, rather than decisive victory, shifts the game: from impossible endpoints to achievable positioning, from nebulous outcomes to actionable systems, and from big data to intentional data. This reframing would require a fundamentally different posture than the one that gave the United States decades of war with no clear definitions of success, particularly in the cognitive domain.
Erol Yayboke is a senior fellow (non-resident) with the Futures Lab at the Center for Strategic and International Studies in Washington, D.C. He is also the chief operating officer at FilterLabs.AI. Nickolas Wilcox is a director at CACI International Inc.
Commentary by Erol Yayboke — September 19, 2025
Commentary by Benjamin Jensen, Nico Vacca, and Jose M. Macias III — May 5, 2026
Report by Jose M. Macias III and Nico Vacca — May 4, 2026
