I think if LLMs improved or our usages of them improved to the point we became design/code reviewers full time many of us would leave to do something less boring and so in some ways there is a negative incentive to investigate different AI driven workflows.