One of the most consistent gaps I observe when working with military leaders on UAS integration is the disconnect between what AI-enabled systems are marketed to do and what they actually do in degraded, contested, or tactically complex environments. This gap creates two failure modes: over-reliance on autonomous capability that does not perform as expected, and under-utilization of genuine capability because leadership does not trust or understand it.

Both are preventable with appropriate training. Neither requires a technical degree to address.

What AI-Enabled UAS Actually Does

Modern edge AI systems — like the NVIDIA Thor AGX in Group 3 platforms — perform inference on-board the aircraft without relying on a ground-based processing pipeline. In practice this means a few specific things:

Pattern recognition at sensor speed. A properly trained model running on-board can identify and classify objects of interest — vehicles, personnel, structures — faster and more consistently than a human operator watching a video feed over hours of flight time. The system does not get fatigued.

Closed-loop tracking. The aircraft can maintain track on a moving target of interest autonomously, adjusting heading and altitude to keep the target centered, without continuous operator input. The operator defines the task; the system executes it.

Autonomous mission execution. With appropriate waypoint programming and decision logic, the platform can execute a reconnaissance mission, identify targets of interest, and return a structured report — without streaming raw video at any point. This matters significantly for operational security in environments where datalinks may be exploited.

What It Does Not Do

It does not make targeting decisions. Current AI systems classify and flag. Engagement authority remains with human commanders and operators. The technology does not change the legal or ethical framework — it changes the speed and consistency of the information provided to decision-makers.

It does not perform well outside its training distribution. A model trained on one environment, season, or sensor configuration may not perform reliably in a substantially different environment. Understanding this limitation is essential for setting realistic expectations.

It does not replace operator judgment. Autonomous systems fail in novel situations. Operators who understand what the system is doing — and what conditions cause it to fail — are more effective than operators who treat the system as a black box.

What Commanders Need to Know

The FFA-201 course (AI Fundamentals for Non-Technical Personnel) is designed specifically for this audience — commanders, staff officers, and senior NCOs who need to make decisions about autonomous systems without needing to understand the underlying engineering. The two-day course covers:

  • What current AI/ML systems can and cannot do in operational environments
  • How to evaluate vendor claims against operational reality
  • Decision authority frameworks for autonomous systems
  • Integration of AI-enabled UAS with existing ISR and targeting processes
  • Ethical and legal considerations — the questions your JAG will ask

The course requires no technical background. It is designed for leaders, not engineers.

View FFA-201 Course Details → | Request Training Proposal →

Mike is the founder of Forge and Flight Academy and a retired U.S. Army Special Operations veteran with 10 deployments and extensive UAS operational experience.