There are algorithms for combining many imperfect models into something much closer to reality than the best model - 'wisdom of crowd' effects etc. There are definitely human organisations that come up with smarter plans than could be come up with by the smartest human in those particular organisations.

If you're talking about AI with 'smarter than us the way we're smarter than chickens' as the standard I suppose I'd have to agree, but it seems like there's a big, very vaguely defined stretch between 'very smart human' and 'makes a very smart human look like an imbecile'