Sure, there's a person turning the key, just like there's full auto on ship defense systems. I can easily see a situation some years down the road where unmanned systems are launched downtown, and before entering a full on denial environment, are told to target fixed position X or column of trucks Y or equipment Z.
All this assumes advances in AI, decision making systems, etc., but isn't terribly far out of the realm of current day operations.
The thing is, robots are dumb.
"Advances in AI" is easy to say but it's light-years away from where we actually are. And even when or if the technology does get there, there still would have to be the legal and ethical and political will to let it be used, and to take responsibility for what a robot decides to kill. Maybe I'm just cynical, but I can not see any time in the foreseeable future envision a theater commander or higher being comfortable with that.
CIWS is never set to auto-engage unless you're in a real-time, holy shit,
Red Storm Rising-style shitstorm of incoming. CRAM - CIWS on a truck for knocking down mortars - has limited range and there's a lot of cutouts in place to ensure nothing gets shot at unless it's really incoming. I've seen a CIWS go nuts trying to lock on to spinning helo rotor blades because they didn't put it in standby before flight quarters. At least one of the blue-on-blue Patriot incidents during OIF 1 was because the battery was in auto-engage mode after an air raid warning forced the crew into bunkers. It met the engagement logic but there was no human in the loop to say, "wait, I don't think that's really an Iraqi jet because that doesn't make any fucking sense".
At most, I can see maybe a consent-to-engage logic where it tells the AVO/MC "this looks like what you asked me to kill. Can I kill it?"
UAS has value for ISR and targeting things where a long-stare is needed (who's in that house? is that our target in that truck?) in order to satisfy ROE. What you're talking about - going after fixed or clearly identifiable targets - we can already do with conventional missiles. Why send a drone to do it? They're more expensive and less useful than simply lobbing a Tomahawk or JDAM in. If you need to make an intelligent decision on whether to engage (intelligent in the sense of applying risk assessment and fuzzy logic, not necessarily smart decisions), there will always be a human in the loop.