WHAT AI CAN FIX: THE STRUCTURAL TRAINING DEFICIT
ELECTRONIC WARFARE TRAINING & ARTIFICIAL INTELLIGENCE
The ten shortcomings documented at NTC Fort Irwin are real — and several are exactly the kind of problem AI is well-suited to address. But not all of them. Knowing the difference is the beginning of a viable solution.
THE TRAINING PROBLEM IN PLAIN TERMS
The Army’s ten documented NTC EW training shortcomings, extracted from years of OPFOR observations at Fort Irwin, sort into two fundamentally different categories. The first category is structural and logistical: soldiers are not getting enough exposure to electromagnetic warfare effects because of regulatory constraints on jammers at home-station installations, because doctrinal materials update too slowly to reflect the current battlefield, and because EW systems are fragmented enough that no individual soldier or commander can see the whole picture. The second category is cognitive and cultural: commanders freeze when their common operating picture disappears, trust between combat arms and EW specialists is thin, and the instinct to wait for the network to come back rather than fight through degraded conditions is deeply ingrained.
These two categories respond to AI very differently. The structural problems — too little training repetition, too slow a doctrinal update cycle, too little battlefield context reaching the schoolhouse — are precisely the kind of problem that AI, properly applied, can solve at scale and speed. The cognitive and cultural problems require AI as an enabler but fundamentally demand human institutional change that no algorithm can mandate. Confusing the two leads to either overconfidence in what technology can fix, or dismissal of genuinely powerful tools because they cannot fix everything.
WHAT AI CAN FIX: THE STRUCTURAL TRAINING DEFICIT
The most debilitating NTC finding is also the most fixable: the “first-jam-at-Fort-Irwin problem.” Maj. Gen. Paul Stanton put it plainly in 2023: the first time most soldiers experience electromagnetic jamming should not be during a graded NTC rotation. The reason it happens that way is regulatory, not pedagogical. Home-station installations cannot freely activate GPS and cellular jammers without FAA and FCC coordination that adds weeks of administrative friction. Even after the Joint Special Warfare Center and School formally requested expanded EW testing ranges in December 2025, the regulatory environment had not changed.
AI-enabled simulation dissolves this constraint entirely. The Army’s Synthetic Training Environment (STE), already in deployment at combat training centers and now expanding to home-station installations, can replicate electromagnetic jamming effects with high fidelity in a virtual environment that requires no RF emissions at all. As Army Futures Command STE program lead Frank Tucker explained, the eBullet component of STE can simulate the propagation of radio-frequency jamming — “we’ve never been able to train this stuff, never” — using effects models rather than physical RF energy. A soldier running a degraded-comms drill in an STE virtual environment on a Tuesday afternoon at Fort Cavazos is exposed to the same cognitive stress as a soldier jammed by the OPFOR at Fort Irwin, without a single watt of RF emission.
The STE is not hypothetical. Reconfigurable Virtual Collective Trainers — hardware that connects to the STE using heads-up displays, high-resolution monitors, and controllers — have already been installed at Fort Moore, Georgia, and Fort Cavazos, Texas, providing collective training capability for Abrams, Bradley, and Stryker formations. PEO STRI is slated to begin fielding AI/ML-enabled live training systems at the National Training Center in fiscal year 2026, followed by home-station installations. These systems provide real-time data-driven feedback and adapt training scenarios to individual skill levels. The AI-enabled live training system can, in the Army’s own description, “support small-unit training without the need of trainers, exercise control, OPFOR, or external enablers” — exactly the home-station self-service model that Stanton called for.
“We’ve never been able to train this stuff — never. The Army can now dream up a new weapons system and synthetically deploy it in a training exercise without bending metal.”
— Frank Tucker, STE Program Lead, Army Futures Command
THE DOCTRINAL UPDATE PROBLEM: ARC AND THE TRAINING-TO-BATTLEFIELD LOOP
The second structural failure is slower and less visible: training materials reflect the battlefield of three to five years ago rather than the battlefield of today. The canonical demonstration is Ukraine. By late 2024, Russian forces had compressed their EW countermeasure adaptation cycle to 48–72 hours — the time between a Ukrainian unit’s first use of a new technique and Russia’s fielding of a countermeasure. U.S. Army training doctrine, updated through traditional curriculum review processes, operates on timelines measured in years. The gap is not a failure of intent but a structural mismatch between the pace of the battlefield and the pace of institutional bureaucracy.AI-Assisted Revisions for Curricula (ARC), developed by the Army’s Artificial Intelligence Research Center of Excellence for Education (AIRCOEE), directly attacks this mismatch. ARC automatically flags outdated content in training materials and generates update recommendations. It was originally piloted at two installations; demand was sufficient that it is now in use at seven Army training centers, including the Fires Center of Excellence at Fort Sill. The parallel Training Development Tool, deployed in 2024, reduced doctrinal revision cycles from years to hours according to testimony before the Army Vice Chief of Staff in November 2024.
For EW specifically, the integration of Ukraine battlefield intelligence into ARC-driven curriculum updates would create a loop that did not previously exist: TRADOC G-2’s OPFOR Modernization Branch tracks adversary EW developments; that intelligence flows into ARC; ARC flags outdated countermeasure guidance in training materials and generates replacement content; updated content flows into PAL3 for delivery to soldiers. The Air Force’s 350th Spectrum Warfare Wing — responsible for rapid EW reprogramming — has already compressed its own waveform update cycle by moving from a centralized lab model to a push architecture that gets updated countermeasures to deployed warfighters faster. The same logic applied to training content could make the schoolhouse nearly as current as the battlefield.
CamoGPT, the Army’s large language model now running on both NIPR and SIPR networks with approximately 75,000 users, already incorporates EW course content in the Cyber Center of Excellence workspace. The CCoE workspace gives CamoGPT access to EW doctrine, lessons learned, and tactics, techniques, and procedures, allowing soldiers and EW planners to query it for current guidance on signal reconnaissance, jamming operations, and spectrum management. The critical next step — not yet implemented at scale — is feeding Ukraine-derived battlefield data into that knowledge base continuously rather than episodically, so that CamoGPT’s EW answers reflect the current tactical environment rather than the doctrinal baseline at the time of the last manual update.
PAL3 AND THE EXPERT DISTRIBUTION PROBLEM
NTC Finding 9 — EW expertise too thin and too senior, impossible to replicate across the force — is one of the strongest use cases for AI-assisted instruction. The 11th ACR’s Capt. Jake Thomas, with seven years of OPFOR EW experience, represents exactly the kind of deep institutional knowledge that cannot be cloned across several hundred thousand soldiers. PAL3, the Army’s Personal Assistant for Life Long Learning developed by AIRCOEE, is specifically designed to distribute exactly that kind of expert knowledge through adaptive one-on-one dialogue.
PAL3 is not a reference tool. It is a Socratic tutor that engages soldiers in interactive dialogue, adapts to their demonstrated knowledge level, and pushes back when their answers reflect misunderstanding rather than confirming whatever they say. Originally used to maintain electronics-technician skills in Navy sailors, it is now deployed on both Google Play and the Apple App Store and has been piloted in Army AI literacy programs with more than 200 students reporting significant learning improvements in a 2024 USC pilot. Applied to EW content — signal recognition, PACE plan execution, electromagnetic signature management, drone RF vulnerability — PAL3 could provide every soldier in a brigade combat team with personalized, adaptive coaching in the specific EW gaps that NTC observations reveal they consistently fail to address.
The critical integration requirement is content: PAL3 is only as current as the material fed into it. The ARC-to-PAL3 pipeline described above — ARC flags outdated content, generates replacements, uploads to PAL3 for adaptive delivery — is the mechanism that makes both tools operationally relevant rather than institutionally static. The Army has the components; it has not yet fully integrated them into a closed-loop system.
COGNITIVE EW AT THE EDGE: THE MACHINE-SPEED TRAINING COMPLEMENT
Beyond training, a second AI application addresses the NTC failures directly in the operational environment rather than the schoolhouse: cognitive EW systems that reduce the demand on individual operator knowledge by embedding AI-driven signal recognition and countermeasure recommendation at the point of employment.
HII’s Cognitive EW platform and DARPA’s PROWESS program both aim at the same fundamental problem: human operators cannot process the speed and complexity of a contested electromagnetic environment fast enough to respond before the window of opportunity closes. Cognitive EW deploys machine learning algorithms at the edge — on the TLS Manpack, on software-defined radios, on drone payloads — to perform real-time signal recognition, RF fingerprinting, and adaptive countermeasure generation autonomously. The PROWESS processors under development by SRI International and USC reconfigure within 50 nanoseconds, a response time no human can match.
The training implication is significant. If the operational system autonomously identifies jamming, classifies the threat, assesses its impact on friendly communications, and recommends or executes countermeasures without waiting for operator input, the floor on what an individual soldier needs to know to function effectively is substantially lowered. This does not eliminate the need for EW specialists — someone must configure, validate, and maintain the system, and someone must interpret its outputs and integrate them into the tactical decision — but it transforms the cognitive demand on the non-specialist from “understand EW deeply enough to respond under pressure” to “understand EW well enough to validate AI recommendations.” That is a meaningfully lower bar for the infantry captain or armor platoon leader whom the NTC consistently shows is the weakest link in the EW chain.
The AOC 2025 symposium’s technical track, themed around “Charting a Path to 2035,” featured multiple sessions on generative AI applied to EW, including real-time cognitive EW using generative AI and edge computing, automation of electronic attack generation with AI, and generative AI for radar and communications EW applications. The convergence of these capabilities with the Army’s STE and training AI stack points toward a near-future training environment where soldiers practice alongside AI teammates that model the cognitive EW behavior they will encounter on a real system — building the human-machine teaming habits in training that will be required in combat.
“AI/ML will be an enormous advantage. Current EW systems and tactics cannot quickly adapt to new emerging threats because of an overreliance on database libraries with predefined countermeasures. Human cognition limits military warfighting capability.”
— Lt. Cmdr. Brian Gannon, USN — USNI Proceedings, August 2023
WHAT AI CANNOT FIX: THE CULTURAL AND COGNITIVE FAILURES
Three of the ten NTC findings resist AI solutions in ways that are important to understand clearly. NTC Finding 1 (digital dependency collapse), Finding 2 (cognitive paralysis without COP), and Finding 4 (the doctrinal-cultural gap between combat arms and EW) are human behavioral and cultural problems. AI can inform them, support them, and create the conditions for change — but it cannot mandate the institutional transformation they require.
The digital dependency failure is not a knowledge failure. The NTC repeatedly shows that Blue Force commanders know they should transition down the PACE hierarchy when jammed. They do not do it promptly, decisively, or without loss of operational coherence — not because they lack the information, but because the behavior has not been drilled to the point of automaticity. This is the same reason the Army still sends soldiers to physical training every morning rather than issuing them a fitness manual: knowledge of what to do is not the same as the conditioned reflex to do it under stress. PAL3 can teach soldiers what a jammed PACE transition looks like. STE can expose them to it repeatedly. Neither can replicate the pressure of a live rotation where the grading consequence is real and the OPFOR is actively probing for exactly this failure.
The cultural gap between combat arms and EW is deeper. Capt. Thomas at the NTC described it as “a steep learning curve that can boggle my mind even now,” and he has seven years of specialized OPFOR experience. The gap is not primarily informational — it is professional. Combat arms officers are socialized to measure competence in terms of maneuver, firepower, and leadership under fire. EW is not part of that identity. Bridging it requires sustained curriculum integration in branch school professional military education, deliberate officer assignment patterns that cross-pollinate combat arms and CEMA, and senior leader emphasis that makes electromagnetic competence a promotion-relevant skill. AI tools can inform all of these efforts. None of them can be delegated to an algorithm.
The Army’s EW Board of Directors — the three-star forum under the G-3/5/7 established in 2025 — and the Cyber School’s partnership with the Maneuver Center of Excellence are the right structural moves for the cultural problem. So is the proposed Electronic Warfare Center of Excellence directed by the FY2025 NDAA, which would give EW its own institutional home in TRADOC rather than subordinating it to the cyber branch. These are human institutional decisions. AI can support the education they enable. It cannot make them.
THE INTEGRATION GAP: THE MOST ACTIONABLE FINDING
The Army already possesses most of the AI tools needed to address its structural EW training deficit. STE provides the jamming simulation environment. ARC provides the curriculum update engine. PAL3 provides the adaptive individual instruction capability. CamoGPT provides the EW knowledge base query interface. Cognitive EW systems provide the operational AI complement that reduces individual operator knowledge demands. DARPA PROWESS provides the edge-computing speed layer. None of these tools are missing from the Army’s portfolio.
What is missing is integration. ARC does not automatically pull Ukraine battlefield intelligence from TRADOC G-2 OPFOR analysis. PAL3’s EW content is not on a regular update cycle tied to ARC outputs. CamoGPT’s CCoE workspace is populated with doctrinal baselines but not continuously updated with operational lessons. STE’s electromagnetic warfare simulation capabilities are being fielded at combat training centers but are not yet universally available at home-station installations in the form that eliminates the first-jam problem. Cognitive EW operational systems and STE training environments are not yet linked so that soldiers practice with AI teammates that model the actual behavior of the systems they will employ.
This is a program management and architecture problem more than a technology problem. Col. Leslie Gorman, Army Capability Manager for EW at the Development Integration Directorate, has described the overarching goal as “owning the spectrum” — analogous to the Army owning the night after the widespread adoption of night vision in the 1980s. That analogy is instructive. Night vision did not succeed because the goggles were invented; it succeeded because the Army built an entire doctrine, training system, and leader development culture around a new way of fighting. AI-enabled EW training faces exactly the same integration challenge: the components exist, and the question is whether the Army’s institutional machinery can assemble them into a coherent system at the pace the threat demands.
A PROPOSED ARCHITECTURE: CLOSING THE LOOP
The following framework describes an integrated AI-enabled EW training system that the Army could build from existing components, requiring program management will and architecture decisions rather than new invention:
At the input layer, TRADOC G-2 OPFOR analysis of Ukraine and other active conflicts feeds continuously into ARC’s content flagging engine. ARC identifies outdated EW doctrine, countermeasure guidance, and tactics across all training materials. The Training Development Tool generates updated content on a near-daily cycle rather than annually. Updated content is reviewed by EW subject matter experts (human review is preserved; AI accelerates the cycle, not bypasses it), then uploaded into PAL3 and the CamoGPT CCoE workspace.
At the instruction layer, PAL3 delivers adaptive individual EW instruction to every soldier in a brigade combat team, not just EW specialists. The tutor identifies each soldier’s specific gaps — PACE plan execution, drone RF signature awareness, deepfake order authentication — and provides targeted remediation before NTC rotation. STE provides collective training environments at home station with simulated jamming effects that replicate the OPFOR experience, allowing units to rehearse PACE transitions under stress before their graded exercise.
At the operational layer, cognitive EW systems embedded in TLS Manpack and future MEMSS platforms continuously classify threats and recommend countermeasures, reducing the floor on operator specialist knowledge. These systems’ behavior is mirrored in STE training scenarios, so that soldiers practicing with virtual AI EW teammates are building the human-machine teaming habits that the real system will require.
At the assessment layer, AI monitors performance data from STE exercises and PAL3 interactions to identify unit-level EW gaps, routing that intelligence back to brigade commanders and the NTC OPFOR cadre. The first-jam problem is eliminated not by never jamming at Fort Irwin but by ensuring that every soldier arrives at Fort Irwin having already been jammed — many times, adaptively, at their home station.
“AI/ML will support small-unit training without the need of trainers, exercise control, OPFOR, or external enablers.”
— PEO STRI, Army AL&T Magazine, July 2025
THE LIMITS AND THE CAUTION
Three cautions are essential. First, AI-generated training content carries the risk of scaling errors at the speed of automation. A Ukraine lesson incorrectly abstracted by ARC, delivered at scale through PAL3 to tens of thousands of soldiers, can embed a wrong lesson more deeply than a manually produced training error ever could. Human expert review must remain a mandatory gate in the ARC-to-PAL3 pipeline, not an optional quality check. The Army’s Center for Army Lessons Learned guidance on CamoGPT is explicit on this: AI outputs must be reviewed and validated by subject matter experts and should never be blindly trusted. That principle applies even more strongly to EW training content, where errors can be tactically lethal.Second, the Army’s CIO has flagged computational cost as a genuine constraint on AI scaling. The transition from bespoke models like early CamoGPT to commercial enterprise platforms like the Army Enterprise LLM Workspace (powered by AskSage) reflects a recognition that open-ended AI deployment is fiscally unsustainable. An integrated EW training AI architecture must be designed with compute cost in mind, prioritizing the use cases — individual soldier EW skills training, curriculum update flagging — that deliver the highest readiness return per token.
Third, and most important: none of the AI training tools described above addresses the irreducible human problem at the center of the NTC findings. The 11th ACR’s OPFOR wins not primarily because it has better jammers — though it does — but because its junior leaders are trained and trusted to exercise independent judgment when communications fail. That trust is built over years of cultural reinforcement, not delivered through an adaptive tutor. AI can prepare soldiers to know what to do. It cannot build the command climate in which a lieutenant trusts that his captain will support him when he acts without orders in a degraded-comms environment. That is the work of the institution, and it cannot be automated.
CONCLUSION
The answer to the question — can AI systems help the Army’s EW training problems by embedding the latest battlefield lessons? — is yes, specifically, and no, broadly. AI is the right solution to the structural training deficit: the first-jam problem, the doctrinal staleness problem, the expertise distribution problem. It is a powerful but insufficient enabler for the cognitive and cultural problems. The Army has the tools. The gap is integration, sustained investment in the ARC-to-PAL3 pipeline, and the institutional will to make electromagnetic competence a measured, graded, promotion-relevant skill rather than a niche specialty left to the CEMA branch.
Ukraine has demonstrated, at operational scale and with lethal consequences, that the Army’s NTC findings are not training artifacts. They are real. Russia’s use of AI-based spectrum analysis tools to detect and jam Ukrainian communications — documented by CLAWS in May 2025 — shows that the adversary is already using AI on the operational side of the EW equation. The question is whether the Army will use AI on the training side fast enough to close the gap before the next large-scale combat operation reveals that the NTC failures were not just training problems after all.
SOURCES AND CITATIONS
NTC / OPFOR Training
[1] Freedberg, Sydney J. Jr. "‘California Jammin’: Wargames Show Army’s Electronic Weakness — and a Human Fix." Breaking Defense, February 2025. https://breakingdefense.com/2025/02/california-jammin-wargames-show-armys-electronic-weakness-and-a-human-fix/
[2] Lawrence, Drew F. "Army Trying to Expose Entire Force to Electromagnetic Warfare During Training." DefenseScoop, August 17, 2023. https://defensescoop.com/2023/08/17/army-trying-to-expose-entire-force-to-electromagnetic-warfare-during-training/
[3] Stanton, Maj. Gen. Paul. "Army Expanding Electronic Warfare Training for Every Soldier." AFCEA Signal. https://www.afcea.org/signal-media/cyber-edge/army-expanding-electronic-warfare-training-every-soldier
[4] Tucker, Patrick. "Special Operators Seek Larger Ranges for Electronic Warfare and Drone Development." Defense One, December 18, 2025. https://www.defenseone.com/technology/2025/12/special-operators-seek-expanded-electronic-drone-warfare-test-sites-us/410248/
STE and AI Training Systems
[5] Marino, Cheryl. "Reality Check: AI/ML-Enabled Live Training Systems." Army AL&T Magazine, July 1, 2025. https://www.army.mil/article/286728/reality_check
[6] Fuller, Haley. "Artificial Intelligence Comes to the Ranks." Military.com, November 10, 2025. https://www.military.com/feature/2025/10/31/artificial-intelligence-comes-ranks-next-wave-of-military-training-tech.html
[7] Rozman, 2LT Jeremiah. "The Synthetic Training Environment." AUSA, 2021. https://www.ausa.org/publications/synthetic-training-environment
[8] Tucker, Frank (quoted). "Training and Simulation Update." Battle-Updates.com. https://battle-updates.com/update/training-and-simulation-update-105/
[9] Army.mil. "The Synthetic Training Environment Revolutionizes Sustainment Training." August 20, 2018. https://www.army.mil/article/210105/the_synthetic_training_environment_revolutionizes_sustainment_training
PAL3, ARC, CamoGPT
[10] Brawner, Keith, Ph.D. "Preparing for the Cyber Battlespace." Army AL&T Magazine, August 25, 2025. https://www.army.mil/article/287582/preparing_for_the_cyber_battlespace
[11] Defense News. "From CamoGPT to Life Skills, the Army Is Changing How It Trains Troops." October 28, 2024. https://www.defensenews.com/land/2024/10/16/from-camogpt-to-life-skills-the-army-is-changing-how-it-trains-troops/
[12] Pharathikoune, Soulynamm. "Enhancing Military Operational Effectiveness Through the Integration of CAMO and NIPR GPT." Center for Army Lessons Learned, March 7, 2025. https://www.army.mil/article/283601/enhancing_military_operational_effectiveness_through_the_integration_of_camo_and_nipr_gpt
[13] Army CCoE. "Instructions for Obtaining a CamoGPT Account and Access to CCoE Workspace." https://cybercoe.army.mil/Portals/131/Documents/Instructions%20for%20Obtaining%20a%20CamoGPT%20Account%20and%20access%20to%20CCoE%20Workspace%20plus%20FAQs.pdf
[14] DefenseScoop. "Army CIO Reining In AI Use Cases to Prevent Excess Costs." August 12, 2025. https://defensescoop.com/2025/08/12/army-artificial-intelligence-camogpt-use-cases-cio-leonel-garciga/
[15] "The Cognitive Limits of Warfare: U.S. Army Professional Military Education, AI Integration." September 2025. https://debuglies.com/2025/09/03/the-cognitive-limits-of-warfare/
Cognitive EW and AI-Enabled Operations
[16] Gannon, Lt. Cmdr. Brian P. "Implement AI in Electromagnetic Spectrum Operations." USNI Proceedings, August 2023. https://www.usni.org/magazines/proceedings/2023/august/implement-ai-electromagnetic-spectrum-operations
[17] HII. "Cognitive Electromagnetic Warfare (EW)." 2025. https://hii.com/product/cognitive-electromagnetic-warfare-ew/
[18] Military Embedded Systems. "Generative AI, Cognitive Electronic Warfare to Anchor Technical Track at AOC 2025." https://militaryembedded.com/ai/cognitive-ew/generative-ai-cognitive-electronic-warfare-to-anchor-technical-track-at-aoc-2025
[19] Gorman, Col. Leslie. "Owning the Spectrum: How the Army Is Reinventing Electromagnetic Warfare." FedGov Today, September 11, 2025. https://fedgovtoday.com/fedgov-blogs/owning-the-spectrum-how-the-army-is-reinventing-electromagnetic-warfare
[20] Lawrence, Drew F. "Army Evaluates Several Evolving Electronic Warfare Concepts at Project Convergence." DefenseScoop, April 15, 2025. https://defensescoop.com/2025/04/15/army-project-convergence-electronic-warfare-concepts/
EW Reprogramming Speed and Ukraine Lessons
[21] Lawrence, Drew F. "Air Force Spectrum Warfare Wing Speeds Organic Waveform Development." Breaking Defense, September 24, 2025. https://breakingdefense.com/2025/09/as-ew-proliferates-air-force-spectrum-warfare-wing-speeds-organic-waveform-development/
[22] CLAWS. "Russia-Ukraine War: Lessons from an Electronic Warfare (EW) Perspective." May 31, 2025. https://claws.co.in/russia-ukraine-war-lessons-from-an-electronic-warfare-ew-perspective/
[23] Cambridge Network. "Accelerating the Next Generation of Electromagnetic Warfare." 2025. https://www.cambridgenetwork.co.uk/news/accelerating-next-generation-electromagnetic-warfare
[24] Youngren, 1LT Kai L. "From Front Lines to Factories: Embedding Industry in US Army Units to Accelerate Combat Iteration." MWI West Point, October 28, 2025. https://mwi.westpoint.edu/from-front-lines-to-factories-embedding-industry-in-us-army-units-to-accelerate-combat-iteration/
AI Wargaming and Simulation
[25] DefenseScoop. "Air Force Aiming to Turbocharge Wargaming with AI." November 24, 2025. https://defensescoop.com/2025/11/24/air-force-ai-for-advanced-wargaming-integrated-force-design-rfi/
[26] War on the Rocks. "Military Gaming to Stay Ahead, But Not the Kind You Think." August 29, 2025. https://warontherocks.com/2025/08/military-gaming-to-stay-ahead-but-not-the-kind-you-think/
SIGNAL Magazine is published by AFCEA International • 4400 Fair Lakes Court • Fairfax, Virginia 22033 • www.afcea.org
No comments:
Post a Comment