I'd rather let actual, human artists do that. Using generative AI for the creation phase steals the part of the work they're good at and like most; it's also what defines them, their signature, what they're known for. Even if they're still involved in the process, like the job Owlcat was offering, they're consigned to a less interesting aspect of the process, something like a simple operator. Finally, if the idea is to make the process faster or cheaper, it also means fewer jobs.
Another aspect that is often overlooked is the long-term impact of generative AI. The result of the AI work is put back into the pool of data that's used for the next training. It means that, in the long run, you end up with a big feedback loop where less and less is original work and the resulting set is a bland regurgitation of what humans did in the past.
So, starting using AI for that is a slippery slope; I think it's irresponsible to encourage it in any way. Using tools is fine, this is something else entirely.