Your line of thinking, though (as you noted) easily misconstrued by ideologists, seems quite reasonable and interesting.
I would guess that current AI tech is many orders of magnitude away, competency-wise, from being able to improve on the efficiency and agility of loosely connected networks of generally autonomous humans.
Even so, ignoring such technical limitations, the issues of control and individual agency are incredibly thorny. How are preferences aggregated? And how does a completely centralized system continually update itself to best serve all participants? Perhaps the competing AI systems can campaign every 4 years :)
Perhaps some sort of AI-human hybrid governmental system could offer a more modest path forwards. We are arguably already beginning to take first steps in that direction.
I wonder about those. I have this thought on the back of my head - now that we have ubiquitous light-speed communication and a big surplus of computing power, couldn't doing central planning via algorithms and automated systems offset some of the "inherent limitations" you mention?
(Honest question, I'm not advocating for communist revolution or anything; just looking for viable alternatives to organizing economy by recursive zero-sum games.)