Can someone develop expert-level evaluative judgment in a field without first “grinding through” years of hands-on practice in that field?

As AI systems handle more production tasks, the value of knowledge work shifting from doing to verifying. 

For example, in the recent GDPVal study professionals completed complex projects in a fraction of the time by verifying and editing AI-generated drafts instead of creating from scratch. Domain experts who normally spent ~404 minutes on a task finished in ~109 minutes when reviewing an AI’s work – with no loss in quality.

Is procedural fluency (knowing how to produce the work) a prerequisite for evaluative fluency (knowing how to assess and improve the work), or can we shortcut straight to expert judgment via alternative learning pathways?