I'm not dismissing its usefulness for those scenario's (see my response to Veltoss below). But people tend to way over-estimate what it is capable of.

Generating an office layout? Yeah absolutely, because that's largely based on prior art, no real innovation required. Though as you noted you'll almost certainly need to "steer" the AI because there's so many variables and permutations that it cannot realistically come up with a perfect solution without real intelligence. It'll require iteration from "someone" no matter how advanced it gets.

But AI as it exists right now won't replace let's say your office manager, who would probably be given the responsibility of planning the office layout. Because their job entails making lots of intelligence based judgment calls. That said; given they will get more AI powered tools to do their job there may be fewer jobs available overall because now your office manager at some big office won't need an assistant anymore.

Note I am not saying that AI affecting our economy isn't happening or won't happen. I'm merely saying that any predictions people are making should be met with a heavy amount of doubt, because there is so much misunderstanding out there.