> Sometimes those questions are dumb and can be easily answered by reading the support documents. So we built a bot that ... responds to questions.
But:
> It won’t be right all of the time.
Does this mean the questions aren't dumb, or that the bot is dumber than the questions?