Home : News : Display
Aug. 28, 2026

Constraint Breeds Lethality: Three Models of Military AI Adoption

Branko Ruzic
©2026 Branko Ruzic

ABSTRACT: This article argues that the United States faces an institutional, not technological, adaptation gap in military artificial intelligence (AI). Constraints drive adversaries toward optimization choices that generate greater operational lethality than abundance-driven Western procurement allows. Departing from algorithmic or budget-centric commentary, it compares three procurement architectures (compliance-optimized, battlefield-validated, and systematically extractive) and treats Ukraine as a distinct wartime adoption system rather than a fourth model. Using Russian ZALA Lancet loitering munition strike data, Chinese-language doctrine on dissipative warfare, and US and NATO acquisition documentation, the analysis offers acquisition executives and combatant commanders five reforms achievable within existing authorities.

Keywords: military AI, defense acquisition, autonomous systems, Ukraine, military-civil fusion

 

Open-source trackers document roughly 2,800 Russian Lancet loitering-munition strikes through mid-2024, with each $30,000–$35,000 system producing stark cost asymmetry against destroyed platforms often worth millions.1 Using these strike data, procurement documents, doctrinal statements, and industry reporting, this article develops a comparative framework revealing why constraint-driven innovation—optimizing for deployment speed, cost-per-effect, and battlefield validation—produces greater operational lethality than resource-rich approaches optimized for compliance metrics.

For US practitioners, the key gap is metric choice—procurement privileges capability measures(accuracy and reliability) over operational metrics (cost-per-effect, mean time to field iteration, and production tempo). The article recommends targeted reforms

  • adopt cost-per-effect as a decision metric,

  • expand Other Transaction Authority pathways with ethical gating for rapid prototyping, and

  • authorize controlled battlefield experimentation with allied partners.

These changes would preserve democratic accountability while narrowing the operational adaptation gap.

Strategic Stakes for US Defense Leaders

For US defense leaders, this adaptation gap poses direct operational risk. The Army multidomain operations concept and the joint all-domain command-and-control framework assume technological overmatch and decision superiority. Yet, if adversaries field autonomous systems faster, iterate countermeasures more rapidly, and mass-produce equipment and munitions more effectively, the United States will enter future conflicts with shrinking windows of superiority that erode during extended operations. The 2026 National Defense Strategy reorders Pentagon priorities around homeland defense and the Western Hemisphere, demoting China from the 2022 National Defense Strategy of the United States of America “pacing challenge” framing to second-priority deterrence in the Indo-Pacific.2 The institutional adaptation challenge persists regardless. Whether or not China is the formal pacing threat, current US acquisition processes cannot match the adaptation tempo China demonstrates through military-civil fusion and Russia validates through battlefield iteration—and the new strategic priorities, from contested hemispheric logistics to narco-trafficking interdiction, depend on the same autonomous systems and rapid-adaptation pathways that peer competition requires. Ukraine provides the last opportunity to test institutional reforms under combat conditions before a peer conflict forces emergency adaptation under fire—with the attendant risks of fielding unvalidated systems or ceding tempo advantage to adversaries. The strategic question is whether US defense leaders institutionalize these lessons now through deliberate reform or learn them through operational setbacks that force adaptation without the benefit of allied test beds or peacetime experimentation.

The Lancet Case: Constraint-Driven Development in Practice

By mid-2024, Russia claimed a Lancet hit rate of 77.7 percent against Ukrainian targets (self-propelled artillery, counter-battery radars, and air-defense launchers), often costing several million dollars each.3 More striking still is that these systems were produced under sanctions, with limited access to advanced manufacturing, and used smuggled Western components, including NVIDIA Jetson modules, for autonomous targeting.4

Operational employment continued through early 2026. In mid-January 2026, a Lancet destroyed a German Gepard anti-aircraft self-propelled gun in the Kharkiv region after reconnaissance UAVs transmitted target coordinates from a ZALA Z-16 all-weather reconnaissance aircraft platform.5 Examples analyzed through early 2026 revealed continuous upgrades, including enhanced antenna systems for extended range, more powerful warheads, and improved video transmission.6 Export variants unveiled at the Dubai Airshow 2025 feature doubled endurance, with the Lancet Item 51E loitering munition reaching 45 kilometers and 50 minutes of flight time.7 The ZALA Aero Z-16E reconnaissance UAV now integrates machine-vision algorithms, enabling automatic target detection, identification, and real-time tracking within the Lancet-E ecosystem.8

This development is not innovation despite constraint. It is innovation driven by constraint, and it reveals a fundamental challenge to prevailing assumptions about military AI advantage.

Not all constraints produce innovation. Resource scarcity can degrade capability when it prevents access to critical technologies or talent, but operational constraints that force optimization toward deployment speed and cost-effectiveness (rather than resource constraints that merely limit access to inputs) can drive superior battlefield adaptation. Western militaries assume superiority flows from algorithmic sophistication and ethical governance frameworks. Ukraine’s battlefield laboratory demonstrates otherwise. The optimization choices that constraints forced on Russia prove operationally superior to the choices that abundance allowed the West to make. Understanding these optimization patterns exposes critical vulnerabilities in US acquisition processes and identifies specific institutional reforms needed to close the adaptation gap before the next conflict validates adversary approaches.

Three Models of Military AI Adoption

The Russia-Ukraine War has generated three observable models of military AI integration, each optimizing for variables while accepting trade-offs. These models represent institutional responses to the challenges of fielding AI capabilities under resource constraints, ethical obligations, and operational pressures. Each reflects a replicable pattern of optimization, not merely a national characteristic. Any military could theoretically adopt any model’s optimization logic. What makes the potential responses strategically significant is that each model’s structural strengths directly exploit another model’s vulnerabilities, creating asymmetries that determine operational outcomes independent of technological sophistication. Analyzing what each model prioritizes and sacrifices reveals why constraint can accelerate innovation while abundance can impede it.

Model One: Compliance-Optimized Acquisition (Western Approach)

“Western” here names a procurement architecture rather than a geographic category—FAR/DFARS-based acquisition (the Federal Acquisition Regulation and the Defense Federal Acquisition Regulation Supplement), NATO Standardization Agreements, audit-based oversight, and democratic accountability mechanisms that, together, constrain how governments and organizations field AI capabilities. The 2026 National Defense Strategy places this architecture under stress without dissolving it.

Western military AI development rests on ethical foundations codified in frameworks like the US Department of Defense Responsible Artificial Intelligence Strategy and Implementation Pathway.9 These frameworks mandate lawful use, human accountability, and strategic stability through requirements for explainability, human-in-the-loop controls, extensive testing, and formal certification. Major defense contractors echo this logic in public posture statements on responsible AI development. From a democratic governance perspective, this approach preserves legitimacy, interoperability among alliances, and compliance with international law.

The optimization choice is clear—prioritize ethical oversight, technical sophistication, and risk minimization. The trade-offs are equally clear—deployment speed, cost-per-effect ratios, and rapid adaptation cycles. Defense programs typically take five to 10 years from concept to deployment.10 Early threat analysis dictates rigid requirements that persist throughout the project life cycle. While multiple review gates successfully minimize risk, this model restricts responsiveness to evolving operational contexts.

The result is exquisite platforms in small numbers, at high costs, with limited tolerance for battlefield experimentation. A single MQ-9 Reaper aircraft costs more than $30 million and delivers unmatched intelligence, surveillance, and reconnaissance capabilities. It is neither expendable nor replaceable at speed.11 Western militaries optimize for capability metrics (accuracy, reliability, and compliance documentation) rather than operational metrics (cost-per-effect, production tempo, and mean time to adapt under adversary countermeasures).

This optimization produces specific institutional vulnerabilities. Procurement systems structured around risk elimination before deployment cannot easily incorporate battlefield feedback during development. Programs designed for peacetime acquisition timelines struggle to keep pace with the adaptation speeds modern conflicts demand. Metrics optimized for laboratory validation correlate poorly with combat effectiveness when adversaries deploy countermeasures that controlled testing environments cannot replicate.

Recent developments suggest that institutional recognition of this gap is underway. In July 2025, Ukraine formalized its role as a defense test bed through the “Test in Ukraine” initiative, which invited Western manufacturers to validate systems under live-fire conditions.12 United Kingdom Minister for Defence Readiness and Industry Luke Pollard described Ukraine as “a laboratory for war,” calling for procurement timelines to compress from six years to two and reporting British drone deliveries to Ukraine had grown tenfold in a single year, from 10,000 in 2024 to 100,000 in 2025.13 German contractors now push software updates directly into theater based on combat feedback.14 North Atlantic Treaty Organization (NATO) doctrine is shifting to incorporate lessons about drone swarms and algorithmic decision support.15 But this formalized experimentation operates outside traditional procurement channels. Quantum Systems, Palantir, and dozens of start-ups validate AI-enabled drones through Ukraine’s Brave1 structured pipeline—not Pentagon programs.16 Patriot missiles intercept Russian Kh-47M2 Kinzhal air-launched ballistic missiles and provide critical feedback to NATO engineers for system refinement, but these are emergency adaptations to an ongoing war, not institutionalized peacetime acquisition reforms.17 The West has created parallel pathways for rapid learning. The strategic question is whether America and her allies can institutionalize these lessons before the next conflict or whether they remain crisis-driven exceptions that activate only under existential pressure.

By early 2025, Ukrainian forces had begun deploying Bumblebee drones funded by former Google CEO Eric Schmidt, featuring AI-powered autonomous terminal-attack capabilities.18 By spring 2025, these systems completed more than 1,000 combat flights, with thousands more since.19 Russian technical intelligence reports analyzing captured Bumblebee drones acknowledged “the highest quality” components and warned that the technology “will demonstrate its effectiveness.”20

Ukrainian Defence Minister Mykhailo Albertovych Fedorov has articulated the endpoint of this development—robotic systems doing all the fighting, kill zones emptying of people entirely, and unmanned systems conducting combat among themselves on the ground and in the air.21 Battlefield-validated AI development is happening, but primarily through entrepreneurial pathways that bypass traditional acquisition, raising the question of whether US institutions can adopt similar agility without crises forcing the changes.

Model Two: Battlefield-Validated Iteration (Russian Approach)

Russian military AI integration operates under fundamentally different constraints than compliance-optimized acquisition. International sanctions restrict access to advanced semiconductors, Western manufacturing expertise, and global supply chains. Yet, these constraints forced Western militaries to make optimization choices they actively avoid (accepting higher technical risk, tolerating battlefield experimentation, and prioritizing rapid deployment over comprehensive testing).

The Lancet program exemplifies this model. Development accelerated after 2022 through continuous operational testing, not laboratory validation. Each combat employment generates performance data that inform the next production run. Upgrades deploy not through formal certification processes but through direct manufacturer-to-operator feedback loops. When operators identify tactical gaps—insufficient range, inadequate explosive yield, vulnerable communications—ZALA aeroengineers respond within weeks.

This iteration velocity creates strategic advantages that laboratory-validated systems cannot match. Adversaries observe Western systems, develop countermeasures, and field responses faster than Western procurement can adapt. Russian engineers can move from observing a radar emission or thermal signature to fielding a counter-modification in weeks rather than months. The adaptation cycle operates inside Western decision timelines.

Cost asymmetry amplifies this advantage. Lancet production occurs at scale because it sacrifices technical sophistication for manufacturability. Each $30,000 strike against multimillion-dollar platforms generates economic attrition Western militaries cannot sustain indefinitely. When Ukrainian forces destroy a Lancet, Russian production replaces it within days. When Russian forces destroy a Western platform, replacement requires months or years of bureaucratic approval, contractor negotiations, and allied coordination.

This model accepts trade-offs Western militaries reject. Technical reliability suffers, with operational security limiting public verification of failure rates. Ethical oversight is minimal. Autonomous target selection algorithms operate without meaningful human review. Supply chain vulnerabilities persist. Western component interdiction disrupts production, forcing continuous workarounds and quality degradation.

Yet, operational tempo compensates for technical limitations. Systems reach battlefields while Western equivalents remain in development. Lessons integrate into doctrine while Western militaries debate requirements. Production scales while Western capacity remains constrained by peacetime industrial-base assumptions.

The strategic implication is stark. Resource scarcity, when coupled with institutional tolerance for operational risk, generates faster adaptation than resource abundance coupled with risk aversion. This model illustrates the constraint-breeds-lethality dynamic—not because poverty produces better technology but because necessity forces optimization choices that prove operationally superior under combat conditions where speed matters more than perfection.

Model Three: Systematically Extractive Integration (Chinese Approach)

Chinese military AI development represents neither the compliance-optimized Western model nor the improvised Russian approach but systematic extraction of strengths from both while avoiding their respective weaknesses. Model three relies on deductive analyses from observable indicators rather than direct documentation of Chinese institutional processes that remain opaque to Western observers. Three publicly observable indicators reveal this systematic approach, suggesting deliberate synthesis rather than simple imitation.

What distinguishes model three from the other two is that China employs Western and Russian approaches simultaneously through structural features neither model can replicate without abandoning core commitments. First, military-civil fusion (军民融合) routes commercial AI development into military use through established channels rather than crisis improvisation—a pathway the Western model deliberately separates, while the Russian model lacks the commercial base to build. Second, the central standards authority empowers the state to set AI specifications directly through integrated civil-military standardization bodies, while Western standards emerge through consortium negotiation, and Russian standards remain ad hoc. Third, sustained state-directed industrial policy treats AI as a strategic priority independent of conflict, while Western procurement reacts to threat assessments and Russian production scales only under wartime necessity. These approaches are not gradations on a spectrum, they are categorically different procurement architectures.

Publicly observable indicators support these institutional features. China maintains access to cutting-edge semiconductor fabrication, AI research networks, and advanced manufacturing infrastructure that Russia lacks, and the West restricts for military use. Chinese military exercises increasingly emphasize rapid prototyping cycles and operational experimentation, practices Western militaries conduct episodically during crises rather than systematically during peacetime. Chinese defense white papers explicitly reference “integrated military-civil fusion” strategies that leverage commercial AI development for military applications at speeds traditional defense contractors cannot match.22

While incomplete, this observable pattern matters strategically because it suggests China is not choosing between Western technological sophistication and Russian adaptation speed—it is pursuing both simultaneously. Commercial partnerships provide technological depth. Military-civil fusion provides deployment speed. Authoritarian governance eliminates democratic accountability constraints that slow Western acquisition.

The 2024 Chinese defense exercises demonstrated capabilities that suggest systematic institutional learning. Autonomous maritime surface-vessel swarms conducted coordinated patrol and interception operations using distributed AI decision making in the L30 unmanned surface vessel exercise off the coast of Zhuhai, China, in March 2026 in a show of technical sophistication matching Western research while demonstrating operational employment concepts Russia developed through combat necessity.23 The Joint Sword-2024B exercise demonstrated China’s capacity for multidomain encirclement operations, coordinating naval, air, coast guard, and rocket forces in simulated port blockades and strike operations against Taiwan.24

Chinese strategic literature theorizes Western AI governance frameworks not as norms to emulate but as exploitable architectural features. Wang Ronghui’s two-part development of “dissipative warfare” (耗散战) in the PLA Daily—first in May 2023, then in a more operationally explicit September 2025 follow-up—sets out a doctrine of intelligentized conflict whose explicit aim is sustained entropy-induction on adversary military, political, and economic systems through a closed loop of sensing, decision making, action, and evaluation.25 Wang’s framework treats Western decision latency—the gap created by human-in-the-loop review, certification gates, and audit-based oversight—as the operational opening. Dissipative warfare operates inside those review cycles, raising adversary entropy faster than adversary procedures can resolve it.26 Chinese planners are not converging toward Western governance norms under international pressure, they are designing around them.

The strategic challenge for US planners is that China combines resource advantages (advanced technology, large-scale manufacturing, and unlimited defense budgets) with institutional advantages (rapid decision-making authority, tolerance for operational risk, and systematic battlefield lesson integration). The West cannot match the Chinese resource advantages through increased spending alone. Defense budgets operate under fiscal constraints and democratic oversight. Russia cannot match Chinese technological sophistication. Sanctions and isolation limit access, but the United States could match Chinese institutional advantages through deliberate acquisition reform—if it chooses to adopt optimization priorities that constraint forced adversaries to discover.

This dilemma brings to light the deeper strategic lesson of model three. The constraint-breeds-lethality dynamic is about necessity forcing organizational choices that abundance allows militaries to avoid, not scarcity producing innovation. China proves these choices are adoptable by resource-rich actors who recognize their strategic value. The question for US defense planners is whether recognition occurs before the next conflict or after combat validates adversary approaches.

The Ukrainian Wartime Adoption Model: A Different Analytic Object

The three procurement models compared above are peacetime frameworks for fielding military capabilities—durable systems built to produce weapons over time under fiscal, legal, and political constraint. Ukraine’s experience since 2022 eschews these models. It rejects a procurement model and instead embraces a wartime adoption system, a set of mechanisms for turning commercial technology into combat capability under existential pressure and on timelines and through pathways no peacetime architecture would permit.

The system has four visible features. First, initiatives such as United24 and Come Back Alive mobilize civilian capital to meet specific battlefield needs within days, bypassing traditional procurement. Second, Brave1 functions as a structured pipeline, moving dual-use technologies (such as drones, AI vision systems, and electronic warfare hardware) into combat use within weeks.27 Third, software moves on wartime cycles, with developers working in close proximity to units and pushing updates under fire.28 Fourth, small firms, volunteer engineers, and uniformed operators act as one innovation system rather than separate commercial and military worlds, deliberately blurring the supply chain.

These features currently make Ukraine the world’s most rapidly adapting military AI system and are not transferable as a peacetime model. The system depends on existential threats to sustain regulatory tolerances, public mobilizations, and acceptance of personal risks that compress adaptation to weeks. Peer-state militaries can study Ukraine and adopt specific mechanisms—the embedded engineer model, structured dual-use pipelines, and contracting tied to combat performances—but they cannot reproduce the underlying conditions without a comparable war.

This wartime adoption system illustrates how Ukraine synthesizes the three-model comparison rather than displacing it. Ukraine is the proof of concept that constraint-driven adaptation produces results Western procurement cannot match through current means. The strategic question, which the recommendations below answer, is which Ukrainian mechanisms can peer-state militaries institutionalize before a crisis forces emergency adoption?

Strategic Implications: Why Optimization Choices Determine Outcomes

The three-model framework reveals a fundamental asymmetry in military AI competition. Technological superiority does not automatically translate to operational dominance when adversaries optimize for different variables. Western militaries excel at producing sophisticated systems. Russian forces excel at fielding adapted systems rapidly. Chinese forces appear positioned to excel at both—not through superior algorithms but through institutional processes that prioritize operational metrics over compliance documentation.

This disparity creates specific vulnerabilities for US defense planning. Current acquisition processes assume superiority flows from capability advantages, such as more accurate targeting, more reliable communications, and more sophisticated algorithms. Yet, Ukraine demonstrates that operational advantages flow from adaptation speed, with faster iteration cycles, more rapid countermeasure deployment, and quicker production scaling. When adversaries adapt faster than US forces can respond, the capability advantages erode continuously throughout the conflict.

The metric mismatch drives this dynamic. Pentagon procurement evaluates systems against laboratory performance standards, for example, accuracy under controlled conditions, reliability across specified scenarios, and explainability for ethical review. These metrics correlate with peacetime acquisition success but poorly predict combat effectiveness. Adversaries evaluate systems against operational standards, including cost-per-effect ratios, mean-time-to-adapt under countermeasures, and production tempo under wartime surges. These metrics matter more when conflicts extend beyond initial engagement assumptions.

Three strategic gaps emerge from this analysis.

  1. The tempo gap. Western acquisition operates on five-to-10-year timelines, while adversaries demonstrate six-to-18-month adaptation cycles.

  2. The cost-asymmetry gap. Western platforms cost millions while adversary systems cost thousands, enabling attrition rates US forces cannot sustain.

The learning gap. Western forces validate systems through controlled testing, while adversaries learn through operational employment, generating feedback controlled environments cannot replicate.

These gaps compound during extended conflicts. Initial US capability advantages erode as adversaries iterate countermeasures faster than acquisition cycles can respond. Meanwhile, cost asymmetries enable adversary production at a scale that peacetime industrial base planning does not assume, while Western systems wait out formal requirement updates that adversaries have already rendered obsolete.

The implication for strategy is direct—the United States will not lose future conflicts because adversaries field superior AI algorithms. It will lose them because adversary institutional processes adapt faster than US bureaucracy can respond, and adversary cost structures enable production scales that US acquisition cannot match under current optimization priorities.

Recommendations: Institutional Reforms for Operational Parity

The preceding analysis identifies specific institutional vulnerabilities that adversary approaches exploit. The following recommendations target those vulnerabilities through reforms achievable within existing authorities and current fiscal constraints. They require institutional will, not congressional authorization or budget increases. Implemented together, they create a decision architecture that preserves democratic accountability and ethical frameworks while matching adversary adaptation tempo and cost effectiveness.

These five recommendations form a sequenced approach. Allied operational test beds (recommendation 1) generate the battlefield performance data required for metric reform (recommendation 2) and outcome-based contracting (recommendation 5). Formalized crisis-response pathways (recommendation 3) provide institutional infrastructure for sustained adaptation. Protecting operational research (recommendation 4) ensures future innovation capacity survives budget constraints. Together, the recommendations enable optimization choices that constraint forced adversaries to adopt through deliberate institutional design rather than desperate improvisation.

Establish Allied Operational Test-Bed Authority

Strategic outcome: Institutionalized battlefield validation enables US forces to enter conflicts with systems proven against adaptive adversaries rather than laboratory-validated platforms encountering real countermeasures for the first time. This competitive edge compresses the learning curve that currently disadvantages forces during initial combat phases when technological surprise matters most. Continuous combat feedback establishes adaptation cycles that match adversary tempo throughout extended conflicts, preventing the capability erosion that occurs when adversaries adapt faster than US systems can respond.

The Department of War should establish formal authority for allied operational test-bed partnerships that validate AI-enabled systems under combat conditions with embedded ethical oversight. This initiative builds on the Defense Innovation Unit’s (DIU) existing pathways for rapid prototyping with allied partners while formalizing battlefield validation as a permanent acquisition requirement rather than a crisis-driven exception. This authority would create standing test-bed agreements with Ukraine and future allied partners experiencing active conflict, providing legal frameworks for rapid prototype validation under fire; embed Department of War representatives with authority to collect operational data, observe battlefield employment, and transmit lessons directly to program managers without waiting for formal assessment cycles; and establish ethical review boards that operate on 72-hour decision timelines, not six-month review processes, while maintaining meaningful human oversight of autonomous systems testing.29

This recommendation addresses the core adaptation gap because Western systems validate through laboratory testing while adversaries learn through combat employment. Laboratory testing optimizes for controlled conditions. Combat employment reveals failure modes that controlled environments cannot replicate, such as adversary countermeasures, cascade failures under operational stress, and cost-effectiveness against adaptive opposition.

Current experimentation through Ukraine’s Test in Ukraine program demonstrates feasibility but operates outside formal Department of War authority. Formalizing this experimentation as permanent institutional capacity would enable systematic lesson integration rather than episodic crisis learning. The ethical oversight requirement ensures democratic accountability without sacrificing the adaptation speed that operational testing enables.

Adopt Cost-Per-Effect as Primary Acquisition Metric

Strategic outcome: Adopting cost-per-effect as a primary metric enables force structure decisions that prioritize operational sustainability over platform exquisiteness. This approach addresses the strategic vulnerability adversaries exploit through cost-imposing strategies, forcing the United States to expend expensive platforms against cheap munitions until inventory constraints limit operational freedom. Cost-per-effect metrics reveal when technical sophistication provides diminishing operational returns, enabling program managers to optimize for mass production of adequate capabilities rather than boutique production of exquisite platforms that cannot be replaced at operational tempo.

Program managers should report cost-per-effect ratios alongside traditional capability metrics for all AI-enabled systems. Cost-per-effect measures total program cost divided by confirmed battlefield effects; for example, targets destroyed, missions accomplished, and adversary capabilities degraded. This metric would force explicit trade-offs between technical sophistication and operational affordability. A $30 million Reaper that destroys 100 targets delivers $300,000 cost-per-effect, whereas a $30,000 loitering munition that destroys one target delivers $30,000 cost-per-effect, which is 10 times more cost-effective despite vastly inferior technical capability. The metric would also reveal platform scarcity risks. Systems optimized for maximum capability but produced in limited numbers create single-point-of-failure vulnerabilities, while systems optimized for adequate capability but produced at scale enable attrition tolerance that extended conflicts require. Finally, the annual cost-per-effect calculation reveals whether programs maintain operational relevance as adversaries adapt; systems whose cost-per-effect degrades over time signal that adversary countermeasures are evolving faster than US systems adapt.

This recommendation directly addresses the metric mismatch driving current vulnerabilities. Capability metrics (accuracy, reliability, and range) measure what systems can do in ideal conditions. Cost-per-effect measures systems’ accomplishments against adaptive adversaries under resource constraints. The latter predicts combat effectiveness better than the former when conflicts extend beyond initial assumptions.

Implementation requires operational data that allied test-bed authority systematically generates. Without battlefield measurements, cost-per-effect remains theoretical. This recommendation depends on adopting test-bed authority first, demonstrating how the sequenced approach creates mutually reinforcing capabilities.

Formalize Crisis-Response Acquisition Pathways as Permanent Structures

Strategic outcome: Permanent rapid-acquisition pathways eliminate the institutional lag that currently limits adaptation to crisis conditions. This structure ensures US forces begin conflicts with systems designed for the current threat environment rather than systems designed for the environment that existed when requirements were first defined years earlier. Sustained rapid pathways maintain innovation capacity during peacetime that can surge during conflict, rather than the current model, where peacetime stagnation requires emergency reconstitution when conflict demands suddenly materialize.

Programs like Replicator, DIU rapid prototyping, and Test in Ukraine partnerships should transition from temporary crisis responses to permanent institutional structures with dedicated funding, clear authorities, and formalized oversight mechanisms.30 The Replicator Initiative demonstrates the model as 18–24-month timelines from concept to fielding, acceptance of 70–80 percent technical solutions over 100 percent exquisite platforms, and production at scale rather than boutique capabilities. This recommendation makes Replicator’s crisis-driven approach the peacetime norm. These pathways should operate under different timeline constraints than traditional programs with maximum 12-month development cycles, acceptance of higher technical risk in exchange for operational learning, and mandatory battlefield validation requirements before production scaling.

This recommendation acknowledges what recent experience demonstrates—US institutions can innovate rapidly when crisis forces urgency. Defense Innovation Unit prototypes demonstrate 18-month fielding timelines. Replicator targets a 24-month deployment. The Test in Ukraine platform validates systems within six-month combat cycles. The adaptation gap exists because these capabilities activate only under existential pressure and deactivate when the immediate crisis passes.

Formalizing rapid pathways as permanent structures would provide the institutional infrastructure for peacetime adoption of wartime innovation tempo. Dedicated funding prevents budget uncertainty from disrupting multiyear development. Clear authorities eliminate bureaucratic ambiguity that slows decision making. Formalized oversight ensures democratic accountability without requiring the extensive review gates traditional programs impose.31

The timeline constraints matter strategically. A 12-month development cycle will enable adaptation inside adversary response timelines. Systems fielded in 12 months can observe adversary countermeasures and iterate before adversaries field counter-countermeasures. Traditional five-to-10-year timelines guarantee obsolescence before deployment—adversaries adapt faster than acquisition delivers.

Protect Targeted Operational Research Despite Fiscal Constraints

Strategic outcome: Targeted operational research maintains the innovation pipeline that generates future capability advantages while budget constraints limit broad-spectrum basic research. This protection ensures that fiscal pressure does not eliminate the research capacity needed to address emerging capability gaps current conflicts reveal. Operationally focused research concentrates limited resources on validated problems rather than speculative inquiries, accelerating the timeline from research insight to fielded capability.

Given current defense budget realities, the Department of War should require program managers to allocate 1–2 percent of major program budgets as innovation reserves for targeted operational research addressing specific capability gaps identified through allied test-bed validation. These reserves should fund 12–18-month university research programs on precisely defined operational questions like AI decision-making performance under GPS-denied conditions, human factors driving operator adoption of autonomous capabilities, and adversary counter-adaptation patterns against specific system types.

This approach maintains research capacity while shifting emphasis from research toward operationally validated questions. It addresses the risk that budget cuts eliminate future innovation capacity while ensuring remaining research investments target gaps revealed by operational employment rather than theoretically interesting but operationally marginal questions.

The 1–2 percent allocation represents achievable reallocation within existing budgets, not new funding requests. For a $10 billion program, this generates $100 million to $200 million for focused research—sufficient for substantial university partnerships addressing critical gaps. The 12–18-month timeline ensures research delivers actionable results before operational contexts change, unlike multiyear basic research programs that risk addressing yesterday’s questions.32

Expand Outcome-Based Contracting for AI Systems

Strategic outcome: Outcome-based contracting realigns contractor incentives from specification compliance to battlefield effectiveness, addressing the principal-agent problem where contractors optimize for contract success rather than operational success. This contracting model shifts risk from government to industry for systems that meet technical requirements but fail operational employment, while rewarding contractors who deliver systems that exceed specifications but prove operationally effective under combat conditions. The resulting capability portfolios reflect what works in combat rather than what satisfies peacetime evaluation criteria.

The Department of War should authorize outcome-based contracting for AI-enabled systems by paying for confirmed effects (cost-per-effect ratios, demonstrated adaptation cycles, and validated production tempo) rather than deliverables (such as engineering hours, units produced, and capabilities demonstrated in test environments). This approach would align contractor incentives with combat effectiveness rather than specification compliance.

Outcome-based contracts change fundamental incentives. Traditional contracts reward meeting specifications such as delivering the required accuracy, passing the certification tests, and completing the documentation. Outcome-based contracts reward operational success, achieving the cost-per-effect ratio, demonstrating the adaptation speed, and sustaining the production tempo. Contractors optimizing specifications prioritize laboratory performance. Contractors optimizing outcomes prioritize battlefield effectiveness.

Implementation requires operational validation data that allied test-bed authority would systematically generate. Without battlefield performance measurements, outcome-based contracts cannot define success criteria. This recommendation depends on first adopting test-bed authority, then leveraging operational data to enable contracting reform. The sequenced dependencies demonstrate why piecemeal reform fails—these recommendations work together or not at all.

Conclusion: The Window for Adaptation

The West will retain AI technological superiority for the foreseeable future, measured by algorithm sophistication, model capability, and research output. That superiority is not in doubt. What remains uncertain is whether technological superiority translates to battlefield dominance when adversaries optimize for different variables that matter more in actual conflicts.

Adversaries are not attempting to outbuild Western AI models. They are building institutional processes Western acquisition is not designed to counter, such as deployment at speed, iteration under fire, and cost asymmetry at scale. Western strengths—abundant resources, sophisticated governance frameworks, and technical depth—have generated institutional vulnerabilities, including a procurement system designed to eliminate risk that produces acquisition slowness, platform scarcity, and operational risk aversion. Adversary weaknesses, particularly sanctions and isolation from advanced supply chains, forced compensating choices like deployment speed over reliability, pragmatic iteration over certification, and production at scale over boutique capability.

Ukraine’s battlefield laboratory delivers a stark verdict on these trade-offs. The race is not about superior algorithms but faster institutional adaptation to operational realities. Western militaries must accept higher operational risk in controlled channels, develop with partners in active theaters, and measure systems by cost-per-effect alongside technical accuracy. Otherwise, the United States will field superior technology that arrives too slowly, costs too much, and adapts too rigidly to match adversary tempo.

The recommendations presented here preserve Western ethical frameworks and democratic accountability while adopting the optimization choices that constraint forced adversaries to discover

  1. allied operational test beds with ethical oversight,

  2. operational metrics alongside compliance documentation,

  3. formalized rapid pathways with inspector general monitoring,

  4. targeted research addressing validated gaps, and

  5. outcome-based contracts tied to battlefield effects.

These reforms are achievable within the current authorities and fiscal constraints. The Department of War already possesses the building blocks—Replicator demonstrates rapid timelines, the DIU validates nontraditional partnerships, and Test in Ukraine proves battlefield experimentation feasibility—but treats the experiments as crisis exceptions rather than permanent architecture. Converting these temporary adaptations into institutional norms requires will, not congressional action or budget increases.

For combatant commanders, this change means future forces either enter conflict with systems validated against adaptive adversaries or discover failure modes under fire. For service chiefs and acquisition executives, it means measuring success by cost-per-effect and adaptation cycles, not specification compliance—and choosing platforms that can evolve faster than adversaries counter them, not platforms that arrive on schedule against a threat that has since passed.

China is not waiting to see which path the United States chooses. The People’s Liberation Army enters the 2030s with systematic processes for rapid fielding, battlefield adaptation, and production at scale—not because Chinese engineers are superior but because Chinese institutional processes prioritize operational tempo over procedural perfection. The US military enters the same period with superior technology but slower institutional reflexes, unless these reforms occur now.

The window for deliberate choice closes with each procurement cycle that optimizes compliance over combat effectiveness. Institutional inertia is a strategic decision—one that cedes adaptation advantage to adversaries who already understand that speed matters more than perfection. The strategic question is whether the United States will make those decisions deliberately or learn their necessity through defeat.

 
 

Branko Ruzic
Branko Ruzic is a defense analyst whose work on military adaptation, autonomous systems, and AI governance has appeared in the RUSI Journal, The Cyber Defense Review, Defense and Security Studies, Small Wars Journal, and the Irregular Warfare Initiative. He consults with European Ministries of Defense about autonomy risk and AI governance. He holds a bachelor of science degree in computer science with a background in counterterrorism and cryptography. He has combat veteran experience from the Balkans and works across Croatia and the United Kingdom.

 
 

Endnotes

  1. 1. LostArmour Open-Source Tracker, “Lancet,” lostarmour.info, n.d., accessed January 2026, https://lostarmour.info; “Exclusive Report: Russia Launches over 2,800 Lancet Drones Targeting Ukrainian Artillery with 77.7% Hit Rate,” Global Defense News: Army Recognition Group, January 12, 2025, https://www.armyrecognition.com/focus-analysis-conflicts/army/conflicts-in-the-world/russia-ukraine-war-2022/exclusive-report-russia-launches-over-2800-lancet-drones-targeting-ukrainian-artillery-with-77-7-hit-rate; and David Hambling, “Russia Has an Arsenal of New AI Drones Built with Smuggled US Chips,” Forbes, August 8, 2025, https://www.forbes.com/sites/davidhambling/2025/08/08/russia-has-an-arsenal-of-new-ai-drones-built-with-smuggled-us-chips/.
  2. 2. US Department of War (DoW), 2026 National Defense Strategy: Restoring Peace Through Strength for a New Golden Age of America (DoW, 2026), https://media.defense.gov/2026/Jan/23/2003864773/-1/-1/0/2026-NATIONAL-DEFENSE-STRATEGY.pdf. The Department of Defense operates under a secondary Department of War designation per Exec. Order No. 14347, 90 Fed. Reg. 43893 (Sept. 5, 2025).
  3. 3. “Russia Launches.”
  4. 4. Spencer Faragasso, “Russian Lancet-3 Kamikaze Drone Filled with Foreign Parts,” Institute for Science and International Security, December 18, 2023, https://isis-online.org/isis-reports/russian-lancet-3-kamikaze-drone-filled-with-foreign-parts. See also Hambling, “New AI Drones.”
  5. 5. Justin Young et al., “Russian Offensive Campaign Assessment, January 11, 2026,” Critical Threats Project, January 11, 2026, https://www.criticalthreats.org/analysis/russia-offensive-campaign-assessment-january-11-2026. For the ZALA Z-16 reconnaissance UAV’s role in the Lancet kill chain, see Kateryna Stepanenko et al., “Russian Force Generation and Technological Adaptations Update, July 25, 2025,” Institute for the Study of War, July 25, 2025, https://understandingwar.org/research/russia-ukraine/russian-force-generation-and-technological-adaptations-update-july-25-2025/. The strike was publicly documented January 11–12, 2026, by Russian state sources: Sofya Sokolovskaya, “The Ministry of Defense Showed Footage of the Destruction of the ZSU ‘Gepard’ of the Armed Forces of Ukraine in the Kharkiv Region,” Izvestia, January 11, 2026, https://en.iz.ru/en/2022664/2026-01-11/ministry-defense-showed-footage-destruction-zsu-gepard-armed-forces-ukraine-kharkiv-region; and “Lancet Burned a Gepard Self-Propelled Gun of the Ukrainian Armed Forces in the Kharkiv Region,” ZALA Aero, January 12, 2026, https://zala-aero.com/en/news/lanczet-szhyog-samohodnuyu-ustanovku-gepard-vsu-v-harkovskoj-oblasti/.
  6. 6. Yevheniia Hubina,“Ukrainian Intelligence Agency Reveals Inner Components of Russian Lancet and Scalpel Drones,” Ukrainska Pravda, March 23, 2026, https://www.pravda.com.ua/eng/news/2026/03/23/8026748/.
  7. 7. “At Dubai Airshow 2025, Rosoboronexport Will Showcase a Record Number of Full-Scale Pieces of Armament in the History of Its Participation in Foreign Exhibits,” India Strategic, November 15, 2025, https://www.indiastrategic.in/at-dubai-airshow-2025-rosoboronexport-will-showcase-a-record-number-of-full-scale-pieces-of-armament-in-the-history-of-its-participation-in-foreign-exhibitions/. For upgraded Lancet-E specifications, see “Dubai Airshow 2025 – ZALA Unveils New Versions of the Lancet-E Loitering Munition,” EDR Magazine, November 24, 2025.
  8. 8. “Dubai Airshow 2025”; and “ZALA Aero Unveils Upgraded Lancet-E Export Drone at Dubai Airshow,” RuAviation, November 18, 2025.
  9. 9. Kathleen H. Hicks, U.S. Department of Defense Responsible Artificial Intelligence Strategy and Implementation Pathway (Department of Defense Responsible AI Working Council, June 2022), https://media.defense.gov/2024/Oct/26/2003571790/-1/-1/0/2024-06-RAI-STRATEGY-IMPLEMENTATION-PATHWAY.PDF.
  10. 10. Michael J. Sullivan, Defense Acquisitions: Assessments of Selected Weapon Programs, GAO-17-333SP, (Government Accountability Office, March 2017).
  11. 11. “MQ-9 Reaper,” U.S. Air Force, n.d., accessed January 2026, https://www.af.mil/About-Us/Fact-Sheets/Display/Article/104470/mq-9-reaper/.
  12. 12. “Ministry of Digital Transformation Launches a Platform for Testing Technologies of Global Defense-Tech Companies,” Ministry of Digital Transformation of Ukraine, July 18, 2025, https://www.kmu.gov.ua/en/news/mintsyfry-zapuskaiemo-platformu-dlia-testuvannia-tekhnolohii-svitovykh-defense-tech-kompanii.
  13. 13. Luke Pollard MP, “Speech for Long War Conference” (speech, Royal United Services Institute, November 25, 2025), https://www.gov.uk/government/speeches/rt-hon-luke-pollard-mp-minister-for-defence-readiness-and-industry-speech-for-long-war-conference.
  14. 14. Diehl Defence’s IRIS-T air defense systems in Ukraine, with a cumulative hit rate reported above 95 percent, have benefited from continuous software updates based on operational feedback. See “Diehl Presents the Present and Future of Air Defense in Berlin,” Militäraktuell, May 1, 2026, https://militaeraktuell.at/en/diehl-presents-the-present-and-future-of-air-defense-in-berlin/. On the joint-venture model for in-theater service and iteration, see “Rheinmetall Ukrainian Defense Industry LLC,” Rheinmetall, n.d., accessed June 2, 2026, https://www.rheinmetall.com/en/company/subsidiaries/rheinmetall-ukrainian-defense-industry.
  15. 15. “Harnessing Artificial Intelligence: Allied Command Transformation at the Forefront of NATO Innovation,” NATO Allied Command Transformation, April 16, 2025, https://www.act.nato.int/article/harnessing-artificial-intelligence/. For the broader algorithmic-warfare framing, see Dominika Kunertova,“How NATO Can Integrate AI to Prevail in Future Algorithmic Warfare,” Atlantic Council, March 30, 2026, https://www.atlanticcouncil.org/in-depth-research-reports/report/how-nato-can-integrate-ai-to-prevail-in-future-algorithmic-warfare/.
  16. 16. Olena Kryzhanivska, “Brave1: The Engine Behind Ukraine’s Defence-Tech Community,” Ukraine’s Arms Monitor, November 15, 2025, https://ukrainesarmsmonitor.substack.com/p/brave1-the-engine-behind-ukraines.
  17. 17. Tim Martin, “Raytheon Exec: Patriots to Be ‘As Up to Date as Humanly Possible’ amid European Demand,” Breaking Defense Europe, January 13, 2026, https://breakingdefense.com/2026/01/raytheon-exec-patriots-to-be-as-up-to-date-as-humanly-possible-amid-european-demand/.
  18. 18. C. J. Chivers, “The Dawn of the A.I. Drone,” The New York Times Magazine, updated January 5, 2026, https://www.nytimes.com/2025/12/31/magazine/ukraine-ai-drones-war-russia.html.
  19. 19. Chivers, “Dawn of the A.I. Drone.”
  20. 20. Russian Ministry of Defense technical intelligence report on captured Bumblebee systems, cited in Chivers, “Dawn of the A.I. Drone.”
  21. 21. Andrew E. Kramer, “The Global Profile: Enter the Killer Robots: The Ukrainian Forging the Future of Warfare,” The New York Times, updated May 28, 2026, https://www.nytimes.com/2026/05/15/world/europe/mykhailo-fedorov-ukraine-ai.html.
  22. 22. DoW Annual Report to Congress: Military and Security Developments Involving the People’s Republic of China 2025 (DoW, December 23, 2025), https://media.defense.gov/2025/Dec/23/2003849070/-1/-1/1/ANNUAL-REPORT-TO-CONGRESS-MILITARY-AND-SECURITY-DEVELOPMENTS-INVOLVING-THE-PEOPLES-REPUBLIC-OF-CHINA-2025.PDF.
  23. 23. Jérôme Brahy, “China Tests First Autonomous Maritime Drone Swarm to Counter Future US Naval Operations,” Global Defense News: Army Recognition Group, April 2, 2026, https://www.armyrecognition.com/news/navy-news/2026/china-tests-first-autonomous-maritime-drone-swarm-to-counter-future-us-naval-operations.
  24. 24. Dzirhan Mahadzir, “China Targets Taiwan in Major Military Exercise, Pentagon Condemns ‘Irresponsible’ Action,” USNINews, October 14, 2024, https://news.usni.org/2024/10/14/china-targets-taiwan-in-major-military-exercise-pentagon-condemns-irresponsible-action.
  25. 25. Wang Ronghui 王荣辉, “耗散战：智能化战争典型方式” [Dissipative warfare: A typical form of intelligentized warfare], PLA Daily 解放军报, May 9, 2023; and Wang Ronghui 王荣辉, “从消耗战到耗散战—试析智能化战争制胜方式新变革” [From attrition warfare to dissipative warfare: An analysis of the new transformation in the winning mode of intelligentized warfare], PLA Daily (解放军报), September 10, 2025 (translation by the author; compare Tang’s rendering as “Winning Intelligentized Wars” in footnote 28).
  26. 26. K. Tristan Tang, “Dissipative Warfare: The PLA’s Potential New Strategy in the AI Era,” China Brief 25, no. 17 (2025), https://jamestown.org/dissipative-warfare-the-plas-potential-new-strategy-in-the-ai-era/; and “Military Struggle in the Intelligent Domain” (“智能领域军事斗争”), in Science of Military Strategy (战略学) (National Defense University Press, 2020), 179, 262–74.
  27. 27. Jack Watling and Nick Reynolds, Tactical Developments During the Third Year of the Russo–Ukrainian War, (Royal United Services Institute, February 2025).
  28. 28. Watling and Reynolds, Tactical Developments.
  29. 29. Other Transaction Authority for Prototype Projects, 10 U.S.C. § 4022 (2021).
  30. 30. Kelley M. Sayler, DOD Replicator Initiative: Background and Issues for Congress, Congressional Research Service (CRS) Report IF12611 (CRS, updated January 2026), https://www.congress.gov/crs-product/IF12611.
  31. 31. Defense Production Act Title III, 50 U.S.C. §§ 4531–34 (2018); the industrial-base authority is at § 4533.
  32. 32. The Quick Reaction Fund is an Office of the Secretary of Defense program administered by the Rapid Reaction Technology Office, focused on rapid field-testing of technology prototypes. See US Government Accountability Office, Defense Technology Development: Management Process Can Be Strengthened for New Technology Transition Programs, GAO-05-480 (GAO, June 2005).
 
 

Disclaimer: Articles, reviews and replies, review essays, and book reviews published in Parameters are unofficial expressions of opinion. The views and opinions expressed in Parameters are those of the authors and are not necessarily those of the Department of War, the Department of the Army, the US Army War College, or any other agency of the US government. The appearance of external hyperlinks does not constitute endorsement by the Department of War of the linked websites or the information, products, or services contained therein. The Department of War does not exercise any editorial, security, or other control over the information readers may find at these locations.