Before a spray drone ever lifts off, the operator makes seven decisions in a row: draw the field boundary, classify crops and no-spray zones, pick aircraft and terrain settings, set safety buffers, inspect the generated route, adjust it and recheck coverage, then work out refills and timing. Every one has to be right: every field, every day, from scratch. None of them is hard on its own. There are just a lot of them, they depend on each other, and there is no margin. One buffer set wrong can cost an aircraft. The bottleneck is not the flight. It is the twenty minutes before it.

A site can have a dozen ways out and two or three people on duty. The hard part is not deciding what to do. It is being in several places at once. Fixed cameras see where they were mounted and nowhere else. People move, but one person covers one point. Most drone security systems still need one operator per aircraft, which moves the staffing bottleneck into the air rather than removing it. And nearly all of them depend on network uplink and cloud inference, the part most likely to fail exactly when it matters.

Transmission corridors, pipelines, and high structures run through exactly the terrain that has no signal and no access roads. Today that means sending someone up to walk it, or putting a helicopter in the air. One is slow and dangerous, the other is expensive, and both find problems late. The real obstacle for drones here is not flight. It is what happens past line of sight: no link, no feed, no judgment. The aircraft becomes a camera that follows a route, and the finding happens afterward, when a person scrolls through a few thousand frames.