ASI doesn't mean it can magically reverse-infer a digital process from an existing hodge-podge of automation and manual steps.
ASI doesn't mean there are instantly enough telerobotics or sensors to physically automate all processes.
Similarly, even ASI will by definition be non-deterministic and make mistakes. So processes will have to be reengineered to be compatible with it.
Even that is small minded frankly. The economy would not survive a super intelligence.
Can the smartest developer you know build a project correctly without any requirements?
There's a lot of sausage making behind what to build, how to integrate it with other things, who needs to be told what on which other teams, etc.
Even pie in the sky breakthrough ASI isn't going to be able to do all of that, Day 0.
And if we're using a tautology to define ASI as something that can do that on Day 0, then I'd point out that in the entire history of technology there hasn't been a single advancement that wasn't subsequently refined and improved.
Except maybe fire.
If not, how much smarter do you think someone would need to be to do so?
ASI ridiculousness starts from defining it as a do-anything machine. Everything has limits.
Like… I’ve had a TODO that I’ve expanded and worked on for a few years for a personal project. If I had a tireless bot of my own intelligence, it should only have taken about an hour. Human limitations are immense barriers.