To add to ExSp33db1rd's point, it's also practically impossible to design to eliminate human error, because humans must execute the design. The most you can do is move the potential for error around. And if we ever did somehiow have a design which wasn't done by humans, we'd just have substituted AI-error for human error in all likelihood.
Indeed, overdesigning systems "to be safe" is rarely very successful either. Sometimes you have to trust to the training and skill of someone suitably qualified. (And I'm not saying they were not; even the best pilot in the world can however make a mistake. See, "error, human")