Cute, but your whole premise relies on knowing the right questions to ask, which you don't. We just had an entire decade of good interfaces being ruined by poorly conceived anemic "user stories" we don't need to further destroy our HCI for the next century or so
I would argue the opposite - the premise actually accepts that software developers and product owners can't always know how their software will be used by end users.
Besides, HCI will inevitably change because after 30 years of incremental user interface refinement, your average person still struggles to use Excel or Photoshop, but those same users are asking ChatGPT to help them write formulas or asking Gemini to help edit their photos.
I don't accept the premise that the interfaces were ever actually that good - For simple apps users can get around them fine, but for anything moderately complex users have mostly struggled or need training IMO. Blender as an example is an amazing piece of software - but ask any user who has never used it before to draw and render a bird without refering to the documentation (they won't be able to). If we want users to be able to use software like blender without needing to formally train them then we need a totally different approach (which would be great, as I suspect artistic ability and the technical ability to use blender are not necessarily correlated that strongly).
The right questions aren't always known up front. Some of the reasoning for using AI in the first place is to refine a fuzzy idea, so a tool like this can help one to go from fuzzy to concrete, with good guardrails in place to ensure the concrete is truly solid. The in this case, the point is that the components of good user interfaces are already available, and then composed based on user prompts to their exact specifications and frozen for normal usage. Unfreeze and prompt again later to tweak further, etc.
