Barat Patel argues that data will never be inherently AI-ready due to its inevitable quirks and imperfections. Organizations often waste effort trying to make general data universally AI-ready instead of identifying specific AI use cases first. Clear operational use cases dictate the exact data types, features, and quality thresholds required, preventing teams from chasing the wrong preparatory data problems.
Military exercises rarely mirror the realities of actual combat, making models trained strictly on operational exercises unreliable for the next fight. Rather than trying to produce finished, highly tuned models before deployment, defense agencies must focus on the speed and reliability of the end-to-end pipeline. Practicing how quickly organizations can collect real data, retrain, evaluate, and redeploy models determines military advantage.
Current military cybersecurity frameworks, such as the Risk Management Framework, treat all software identically, requiring tedious full package reaccreditation for every change. Patel argues the Department of Defense should separate AI models and data from base software infrastructure. This architectural decoupling allows defense teams to rapidly push frequent model updates to operational edge units without triggering an entire software re-authorization cycle.
Overly prolonged lab test and evaluation cycles prevent AI capabilities from ever experiencing real-world field conditions. Because enemies, environments, and warfighters dynamically alter operational realities, pre-deployment models cannot be tuned to absolute perfection in advance. Teams must bound evaluation to ensuring basic safety and understanding risks, deploying models into the field quickly so real exposure drives adaptation.
Synthetic data cannot substitute for missing operational data because synthetic engines require representative real-world data to anchor simulations. Military sensors differ significantly from commercial sensors in age, angles, and biomes. Rather than serving as an autonomous training shortcut, synthetic data works best as a targeted augmentation tool to stress-test rare, dangerous edge cases and adversarial deceptions that lack live training examples.
To prevent vendor lock-in, defense programs should keep data ownership, data governance, and pipeline architectures strictly in government hands while outsourcing approximately 90% of operational work to commercial industry. Internal government engineers should develop enough capability to maintain intellectual parity with the state of the art, using a modular acquisition framework that rotates competitive industry partners across data labeling, training, and deployment.
Fielded autonomy in modern warfare, as observed in Ukraine, did not begin with autonomous vehicles; it emerged because technicians spent years actively collecting, transmitting, and annotating video and sensor streams from human operators. In contrast, the US military lacks continuous data collection pipelines for machine learning, frequently discarding or siloing exercise data due to storage costs and institutional neglect of boring data infrastructure.