Let's imagine a future where fully autonomous weapons exist. Advanced AI is driving them, and technological advances have enabled the weapons to have better sight (wide spectrum sensors), better range and faster reaction time than humans. They can go where humans can't and endure in environments where humans can't live even for a couple of minutes. Perhaps initially the weapons are simple, kind of like the "sword" from Second Variety (P. K. Dick), using radio emitters to identify friend from foe. They would also need a power supply, and because you can't put an advanced brain of huge computing power on a small platform (laws of physics still apply), the "brain" would be located somewhere shielded, deep underground, perhaps near a volcano to use the thermal energy, but also near water, because of cooling requirements. The brain would use a large capacity datalink to communicate with the weapons. Humans are still in the loop, of course, because the adversary is unfortunately quite adaptable and you need programmers to change the software as battle requirements demand. However, humans have this pesky thing called conscience, and sooner or later, if you make them fight, they will refuse to pull the trigger. So they need to be removed as far as possible from the decision to kill, left to the AI. So a team (quite large in fact) of programmers are working on small, disparate features (image recogniton, data communication, weapons delivery) but they never quite see the big picture themselves. Since what they do is quite important, they are well paid and enjoy a nice lifestyle. There are of course downsides. Refusing to do the work anymore or even talking about it could bring repercussions. So the best strategy is to keep quiet and mind your own business paycheck to paycheck. Meanwhile, old school soldiers become obsolete and the AI does all the messy work.
Sounds like fiction? Perhaps it is.