I Went to Scout AI’s Bootcamp. They’re Training AI for Battlefield Drones

Colby Adcock’s Scout AI just closed a $100 million round to train AI models for military use. That’s a lot of cash for a company that’s been relatively quiet until now. But after visiting their bootcamp, I get why investors are throwing money at it.

Scout is building AI agents that let a single soldier control multiple autonomous vehicles — drones, ground rovers, maybe even boats — in real-time. Think of it as a force multiplier that doesn’t require a PhD in robotics to operate. The idea is to give small units the kind of battlefield awareness and firepower that normally needs a whole squad.

I spent a day at their training ground in the Mojave Desert. It’s not a fancy Silicon Valley office. It’s a dusty patch of land with shipping containers, mesh networks, and a lot of drones buzzing around. Adcock’s team runs live exercises where soldiers wear AR headsets and issue commands by voice or gesture. The AI translates that into movement orders for a fleet of quadcopters and small rovers.

The models are trained on simulated combat scenarios, but they also ingest real data from exercises. Scout claims their agents can adapt to new terrain and threats faster than traditional military systems. That part I believe — most military AI I’ve seen is clunky and requires constant human tweaking. Scout’s demo was smoother than I expected. The drones didn’t collide, they responded to voice commands with minimal lag, and they could autonomously re-route if one went down.

Still, there’s a gap between a controlled demo and actual combat. Adcock acknowledged that during our talk. He said the $100 million will go toward scaling the training pipeline, improving edge inference, and making the system work in GPS-denied environments. That last one is the real challenge. Most commercial drones rely on GPS, and the military knows that’s the first thing an adversary will jam.

Scout isn’t the only player in this space. Anduril, Shield AI, and a dozen startups are chasing similar contracts. But Scout’s focus on individual soldier control — rather than big command centers — sets it apart. Whether that’s enough to win Pentagon dollars remains to be seen.

I left the bootcamp impressed but cautious. The tech is real, the team is sharp, and the funding is there. But war is messy, and AI that works in the desert might choke in a dense city or under electronic attack. Scout has a lot to prove. At least they’re training for it.