The majority of major, targeted hacks (rather than just sweeping the net for vulnerabilities) - aka, the kind of stuff that the US military cares about - involves sending emails or making phone calls and introducing yourself as Bob from IT, and sorry to bother you but there's a problem that we need to discuss with you, but first a couple questions...
They don't need script kiddies, they need social engineers. Question number one in the job interview should be "Is your native language Russian, Chinese, Farsi, Korean or Arabic?" And even as far as the more traditional "hacking" goes, rather than script kiddies they're going to need people who are going to custom analyze a given system and assess it's individual vulnerabilities, people with real in-depth understanding. One would presume that in most cases that the sort of targets that the US military wants to hack are going to keep themselves pretty well patched to common vulnerabilities.
AIs doing hacking? What are you talking about? This is the real world, not Ghost In The Shell.