The inconvenient aspect is that the list of ‘military targets’ changes in part according to how well they can be targeted(and how cheaply).
If you pretty much have to saturation bomb the entire area, WWII style, the list of targets deserving enough is a lot shorter than if if you can just lob a cruise missile or two; and it gets longer still if, with a fair amount of analyst time, you can target down to individual vehicles or buildings.
One can only assume that the list would be longer still if you could automate the tedious analyst stuff that currently hurts throughput and forces you to focus on higher profile targets.
Our current “Yeah, we operate an effectively indefinite and open-ended targeted killing scheme in a laundry list of countries we aren’t even nominally at war with; this no no larger remarkable, yes?” situation, say, is pretty much only possible because fairly granular targeting is an option.
Better targeting doesn’t give you(at least not for long, it may take a little while for an organization to adjust) “same targets as last time, now with 30% less collateral damage!”. It gives you a new set of targets, many that would previously have been unviable.
