field guide

AI-Enabled Mid-Range Strike Drones (Azov Unit Perspective)

An operational view of deep-strike drones: field modification labs, terminal-only AI guidance, and the argument that adaptation speed, not resources, is the real force multiplier.

updated 2026-06-29 autonomy ukraine operational case-study

First Corps Azov’s Unmanned Systems Department operates modified fixed-wing kamikaze drones at 250 km depth against Russian logistics. The key insight: adaptation speed, not resources, wins. Drones in their factory configuration are obsolete within months — unit-level modification labs are the real force multiplier.

Drone platforms and modifications

Base platforms: - Hornet (US-made): tactical UAV with 50 km range in stock configuration - Darts (Ukrainian-made): twin-engine strike drone - Other undisclosed platforms (not yet captured by Russia, so kept secret)

Modifications applied by the unit: - Starlink integration for beyond-line-of-sight communication (previously limited to 50 km by radio link) - Engine/propulsion upgrades to extend range from 50 km to 250 km - AI terminal guidance for the “last mile” (see ukraine-ai-terminal-phase-drone-strikes for the MoD-level architecture) - Alternative comms systems (not Starlink) that remain undisclosed

The unit started testing modifications in early winter, began combat deployment by January/February. That’s a ~3-month cycle from concept to operational use — a speed that traditional defense procurement can’t match.

Operational concept: road denial through persistent hunting

The campaign is continuous, not mission-based. Units are assigned road sections. Drones patrol in “hunting mode” — loitering and engaging targets of opportunity based on intelligence priorities.

Target priority hierarchy: 1. Fuel tankers and fuel logistics (highest impact — fuel is “the blood of war”) 2. Cargo trucks carrying ammunition and supplies 3. Long-haul trailers and rail cargo 4. Individual vehicles near the front (lowest cargo density per kill)

The cascade effect: destroying a fuel tanker at 200 km depth removes multiple tons of fuel, which cascades to FPV drone operations (generators need fuel), vehicle mobility, and logistics throughput. Compare to destroying a single car with two canisters (40L) near the front — same drone cost, 100x the impact.

Result: confirmed fuel shortages in Crimea and Russian-controlled territories.

AI role: terminal guidance and autonomous engagement

AI operates in the “last mile” — terminal approach only. Capabilities: - Target recognition: classifies vehicle type, distinguishes decoys from real targets - Autonomous engagement: can complete the strike without operator input - Multi-drone coordination: one drone observes, another strikes (ISR/strike teaming)

Policy: man-in-the-loop by doctrine. Operator makes the final strike decision. But the system can operate fully autonomously if necessary — this is a policy constraint, not a technical one.

This is consistent with the MoD’s architecture in ukraine-ai-terminal-phase-drone-strikes — AI is a terminal specialist, not a mission manager.

Russian countermeasures (and their ineffectiveness)

Current Russian responses: - Anti-drone nets strung along critical roads deep in rear areas - Shotgun-equipped guards every 50 meters along critical roads - Zebra-stripe paint on trucks to confuse CV models (see ukraine-ai-terminal-phase-drone-strikes for the adversarial ML angle) - Interceptor drones for air defense

None are effective yet. The officer notes they know which direction Russian countermeasures are developing and already have counter-countermeasures prepared.

The real lesson: unit-level innovation labs

The officer’s primary message to the US: “Drones in their basic configuration, right out of the box, is not something that can work.”