When we started building Chatipod, we assumed customers just wanted a fast answer. Turns out that's only half true.
Two things surprised us:
A fast wrong answer makes people angrier than a slow right one. We optimized for speed first and had to unlearn that.
People test the AI on purpose before trusting it — asking something random to see if it'll make something up. If it does, they leave and don't come back to ask the real question.
We're still figuring out exactly where AI should stop and a human should take over. If you want to see where we've landed so far, it's at chatipod.io.
Curious if anyone else building with AI/automation has hit the same "when does the bot hand off" problem.
Top comments (0)