It's a near certainty that I won't get hit by a car tomorrow - people do get hit by a cars but most people don't get hit by cars most of the time. There is a risk (a fraction of percent), but it's acceptable.
It's an even larger certainty that code will have bugs. As far as I've seen, all code will have bugs, as experience shows, even security oriented code carefully made and reviewed by experts tends to have bugs, but it is possible (though not as likely) that a small piece of carefully audited code will be bug-free. That's a bit different than the risk of getting hit by a car; it might be more fair to say that you have a "risk" of being bug-free comparable to the "risk" of winning a lottery.
If smart contracts are at all useful, they most definitely will be useful before we have near-perfect code writing AI's.
It's also possible (and I feel that this will be the case) that in the end the risks will be too large anyway for large adoption, so we'll collectively decide that for most use cases smart contracts are useful only with a "social" mechanism for overriding them, contrary to the current "code is law" policy of Ethereum.