I think an underrated answer is something like "our current institutions and the way they're organized to deploy human intelligence at scale are already much, much more capable than they're given credit for, and so the substantive gain from AI is more marginal than people realize." Reasons for this fact being underrated are a) these institutions are a normal part of life and so taken for granted; b) dissatisfaction with them is very high now, meaning there is little discursive demand for arguments as to how awesomely powerful and effective they are. But if you step back, and put current grievances to the side, what we've built as a species is already so extraordinary, and importantly it works, in classical Hayekian fashion, by integrating together a vast knowledge base drawn from the whole population of individual humans, all of whom have access to mountains of fine-grained "data" on their local circumstances. That's not something AI is yet able to emulate, and many of its bottlenecks are also our preexisting institutions' strengths.