Scenario: Too Close for Comfort
CPL Lee returns from mission and pulls his truck into the motor pool. The vehicle’s onboard sensors retain captured fault codes, environmental exposure, and operating conditions. A low-power, encrypted recorder preserves Lee’s verbal observations locally, but no radio traffic is generated. Human memory and manual entry are removed from the reporting chain at the point of contact.
As the vehicle enters the maintenance bay, adaptive camouflage netting suppresses thermal and radio frequency signatures, but more importantly, it marks the transition from operator-reported sustainment to sensor-verified sustainment. Cameras identify the platform and configuration. Near-field communication (NFC) and radio frequency identification (RFID) sensors pull vehicle history reports, diagnostics, crew identifiers, and maintenance status directly from the vehicle and the maintainers’ wristbands. The system does not ask a Soldier how long a repair takes or what tools were used — it measures it.
These sensor feeds converge on a local artificial intelligence (AI) stack running on an edge server. Fault codes, maintenance history, environmental exposure, voice annotations, previous mission profiles, and even operator usage are fused automatically, generating a ranked fault assessment. Maintainers receive wrench-ready diagnostic guidance on their tablets through low-power mesh or LiFi links, but no data are pushed upward yet.
The diagnosis suggests corrosion in a wheel speed sensor caused by repeated water crossings. The system already knows this vehicle is mission critical for the next day. Instead of a human initiating phone calls or spreadsheets, the sustainment network evaluates regional inventories, identifies recoverable components on battle-damaged platforms, and generates a controlled exchange recommendation. Approval is routed and executed with minimal human mediation; parts are scanned, inventories update automatically, and the vehicle’s readiness state changes in real time across command dashboards.
Throughout the process, sensors continue to collect automatically and continuously. RFID wristbands log man-hours precisely. Tool issuance and part movement are recorded at the moment they occur. When the repair is complete, no one reports success — the system observes it.
Only at a scheduled window does the system communicate beyond the local area. Sensor-aggregated data are compressed and transmitted in a brief, narrowband satellite burst. What moves upward is not raw data or human estimates, but validated sustainment outcomes: readiness restored, parts consumed, labor expended.
At higher echelons, these sensor-derived outputs enable something previously impossible at scale: trend recognition untethered from subjective reporting. AI models ingest maintenance outcomes across units and identify abnormal failure rates tied to specific components and operating conditions. When a vendor’s wheel speed sensors begin failing disproportionately after water exposure, the system detects the trend before it becomes a crisis. Notice-to-maintainer alerts are queued automatically, triggering inspections the next time affected vehicles enter service — without relying on memory or tribal knowledge.
At the strategic level, sustainment leaders no longer ask, “What are units reporting?” They ask, “What are the sensors showing?” Stockage models adjust dynamically, shifting parts forward based on projected operations and sensor-validated demand. Contracting preferences update in response to observed performance, not anecdote.
Across the formation, commanders retain freedom of action not because sustainment became more automated, but because it became more accurate. Sensors replaced the first layer of human reporting, allowing AI to operate on reality instead of approximation. Technology enabled this shift, but sensors made it meaningful.
Current Capabilities
The story above is only a slight embellishment of technological capabilities that exist today. NFC tags are essentially stickers with embedded circuitry that allow for two-way communication and storage of small amounts of data. Mesh communication exists within Army formations already, allowing Android Team Awareness Kit-enabled devices to broadcast reports and other data to all devices on the network. The AI Integration Center (AI2C) has already implemented applications that broadcast data from users or sensors and that automate reporting by leveraging voice-to-text models on existing systems. This style of mesh communication paired with low-cost sensors presents an opportunity for sustainers to automate large swaths of logistics data collection and reporting.
Through a blend of low-cost sensors, dynamic sustainment AI, ruggedized collaboration, and spectrum-disciplined communications, sustainers in the near future will synchronize sustainment at echelon, regenerate combat power, and execute mission command without drawing unwanted attention, preserving and reconstituting combat power while staying hidden in the electromagnetic battlespace. At the center of this vision are low-cost sensors, which transform sustainment data from subjective human inputs into objective, real-time truth, and in doing so unlock the full value of AI. Many of these sensors are already in widespread use across industry to streamline sustainment and maximize profits.
Garbage In, Garbage Out
Sustainment has long been identified as one of the warfighting functions most likely to benefit from the increasing prevalence of AI. The sustainment community has a deep pool of data to train AI models, much of it unclassified and readily accessible. The gap is not data availability, but data quality. Sustainment information is routinely compromised when human judgment, memory, incentives, faulty data entry, and fragmented systems shape what is reported. These inaccuracies are rarely malicious, even when intentional. Reports are adjusted to avoid scrutiny, protect force structure, reconcile accounts, or compensate for incomplete information. Leaders misinterpret data during entry, delay reporting under operational pressure, or make assumptions to fill gaps. Redundant systems that only partially share data compound the problem, fragmenting information across multiple databases. The result is a sustainment enterprise built on estimates rather than observable truth.
This environment is fatal to AI. Algorithms trained on delayed, subjective, or distorted inputs generate recommendations that appear precise but are fundamentally flawed. Efforts to modernize sustainment without addressing how data are collected simply accelerate bad decisions.
Low-cost sensors bypass this problem entirely. By automating data capture at the point of action — reading fuel levels, logging part usage, and tracking man-hours through wearables — sensors remove the first layer of human subjectivity from the reporting chain. They replace approximation with measurement and create the conditions under which sustainment AI can function as intended.
An Aged Model
Low-cost sensors are a viable potential solution but have not been widely adopted for a variety of reasons. Soldiers still manually measure the volume of fuel in a tanker with a wooden stick, report maintenance hours based on self-reporting, and manually maintain supply stocks in a base of operations container. In an era when shoppers at Walmart can find an item’s exact location on their phone, the Army maintains clerks to manage inventory in a model from the 1950s, including annual wall-to-wall inventories (presumably because we steal just as often as we lie).
We offer excuses for our reluctance to adopt new systems and practices: sensors may be insecure; they may not work with our network; they may reduce our ability to operate in a denied, degraded, intermittent, or limited environment; they are not rugged enough, etc. But many of these objections mask a deeper cultural resistance. Commanders worry that too much transparency will expose inefficiencies. Senior leaders fear the political consequences of reporting ugly data. Sustainers themselves are reluctant to automate tasks that form the basis of their military occupational specialty (MOS) identity. We have found reasons to avoid adoption, and we are delaying movement into the intelligence age because of our excuses. The Army is driving forward with AI adoption, but the development of those models relies on accurate, timely data.
The sustainment community must quickly adopt and integrate low-cost sensor technology into our formations. We must tolerate the risk of unflattering reports and accept that certain career fields will be disproportionately impacted by the automation provided by modern sensor technology. We must adopt these technologies in concert with deliberate AI development efforts to ensure that the data provided integrate with our current automation systems and with the data requirements of the AI developers.
A Need for Leadership and Risk Tolerance
Sensor technology is not sufficient to bridge the gap. Commanders must assess the appropriate level of risk tolerance for sustainment planning. Logisticians have long espoused that civilian logistics models translate poorly to conflict; the impact of a late Amazon delivery pales in comparison to the impact of losing the initiative in an assault due to a delayed fuel delivery. Our mistakes are measured in lives lost, and logisticians have long been acutely aware we are the first to be excoriated when a plan fails.
That risk aversion has translated into inefficient models. It has produced archaic processes like wall-to-wall inventories, manual fuel accounting, and gapping mechanics from the parts needed to complete maintenance actions to improve command supply discipline at the cost of efficiency and enabling modernization. In an organization with an annual budget that exceeds $175 billion, we require commanders to directly approve repair components that exceed $500, with no specific means of evaluating the efficacy or efficiency of the maintenance actions those parts enable. We criticize the tactical risks of commercial logistics models but burden ourselves with bureaucratic and administrative process that we inevitably carry with us to conflict. This bureaucratic inertia is precisely what sensors disrupt: they do not ask permission to measure fuel, or hours, or parts usage — they just do it. But until commanders deliberately choose to embrace this disruption, the Army will continue carrying analog habits into a digital fight and will apply industrial processes in the intelligence age.
Sensors can mitigate many of the cultural aversions that drive these processes by automating reporting processes and accountability and better evaluating the actual impact of sustainment investments.
Combined with local AI stacks, sensors can reduce the need for constant reporting traffic by allowing data to be collected, analyzed, and acted upon at the point of need. This would dramatically decrease the unit’s electromagnetic spectrum signature. Higher headquarters would be provided with sustainment needs instead of subordinate data. The result would be a sustainment system that operated faster and with greater autonomy, remained far harder for adversaries to detect, target, or disrupt in the electromagnetic battlespace, and generated redundancies when faced with disruptions.
And the power grows when data become predictive. With reliable sensor feeds, AI can forecast tomorrow’s failures, adjust stockage levels in real time, and even preposition units based on projected operations rather than static estimates. When a systemic fault is identified — like a vendor producing subpar wheel speed sensors — notice-to-maintainer messages can be generated automatically, alerting every impacted vehicle as it enters service. Instead of reacting to shortages, commanders would be given foresight and informed decision points. Sustainment is a source of operational advantage, not a constraint.
More important, accurate data change the commander’s decision space. With sensor-fed sustainment AI, commanders could set dynamic risk thresholds, deciding when to cannibalize, when to surge forward logistics, when to execute a repair, and when to deliberately run equipment to failure. These decisions are not abstract — they shape the operations officer’s scheme of maneuver and the intelligence officer’s assessment of enemy targeting opportunities.
Sustainment AI is not just a logistics tool. It is a warfighting tool. Low-cost sensors can enable this by creating a continuous, trustworthy flow of sustainment data in contact. They can give commanders the confidence to act, not based on estimates, but on the real condition of vehicles, fuel, weapons, and Soldiers at that moment. Sensors are not only an efficiency enhancer — they are a survivability enabler. And in the transition, they can provide the AI development teams the reliably accurate data needed to train models.
A Time for Change
The Contested Logistics Cross-Functional Team conducted a simple but telling experiment: transfer data directly from four M1A2 SEPv3 tanks into a Mission Command Information System (MCIS). Doing so required a special software upgrade, days of hands-on work with each platform, and a carefully configured mesh network — just to pass a handful of fault codes into the Tactical Data Platform. Once there, an AI-enabled MCIS could answer only basic queries about fault frequency and meaning, and at day’s end it generated a short situation report summarizing what was collected. The experiment proved that we can move maintenance data from platform to decision maker, but it also revealed the limits of our current approach: without low-cost sensors automatically generating accurate, high-resolution data at the platform, even our most advanced AI-enabled systems are left with too little to analyze, too late, and at too high a cost in Soldier time. The Army does not lack AI — it lacks the automated data streams required to make AI matter.
The path forward is clear: rapid experimentation, adoption at scale, and tolerance for the cultural disruption sensors will cause. Some MOS roles will change or shrink, but automation can also free sustainers from administrative burdens to focus on the problems AI cannot solve. We cannot afford to wait for perfect technology or perfect data. We must build models, refine sensors, and adapt processes in parallel — and we must do so quickly.
Transformation in Contact (TiC) provides the ideal framework for this shift. TiC is about adapting while under pressure, iterating in the fight rather than waiting for perfect conditions in garrison. Low-cost sensors and sustainment AI models embody this principle: they can be fielded incrementally, validated in real-world conditions, and refined in direct response to the demands of contested logistics environments. By embedding sensors at the points of maintenance, supply, and transportation and coupling them with local AI stacks, units can transform sustainment practices during operations — accelerating decisions, reducing reporting burdens, and minimizing electromagnetic signatures. The more sensors we field, the more quickly the sustainment enterprise adapts under fire, and the sooner sustainment modernization is not deferred to future force design but becomes a living element of TiC, giving commanders the ability to regenerate combat power in contact and experiment with new sustainment concepts in real time.
The good news is that the sustainment enterprise does not need to build this transformation alone. The Army already has organizations designed to help accelerate the adoption of low-cost sensors, integrate data standards, and scale AI tools across the force. The AI2C is postured to provide exactly the assistance sustainers require. AI2C’s partnerships with program executive officers, cross-functional teams, and U.S. Army Combat Capabilities Development Command give sustainers an institutional bridge to technical expertise they do not organically possess.
--------------------
LTC David J. Paddock is a logistics officer and currently serves as the lead capability developer at the Artificial Intelligence Integration Center at Carnegie Mellon University. He holds a master’s degree from Air University. He has served as a maintenance control officer and forward support company executive officer in 1st Brigade Combat Team, 10th Mountain Division, at Fort Drum, and as a commander of a distribution company in 210th Fires Brigade in Camp Casey, Korea. He is also a graduate of Ordnance Basic Officer Leader Course and the Air Command and Staff College.
CPT Jeremy Kilbride serves as an autonomous systems engineer for the Artificial Intelligence Integration Center in Pittsburgh, Pennsylvania. He commissioned in 2018 as an ordnance officer. He holds a Master of Science degree in artificial intelligence engineering from Carnegie Mellon University.
--------------------
This article was published in the winter 2026 issue of Army Sustainment.
RELATED LINKS
The Current issue of Army Sustainment in pdf format
Current Army Sustainment Online Articles
-------------------------------------------------------------------------------------------------
Social Sharing