Using generative AI to replace toil and not the creative human process is fine imo. Even doing something like generating visual things, to me, is OK if it’s driven by real creative intent and doesn’t result in something that looks low quality. But it’s not very simple to get output that you can tweak in fine ways to get predictable changes based on specific creative intent - human language is not descriptive enough to really capture that. “A picture is worth a thousand words” is accurate. You’re also shooting yourself in the foot when you end up with a ton of assets or systems that you don’t have fine control over because you can’t do something simple like tweak a layer of an image because what you got at the end of the day was just a raster output from a black box.