I agree with you that it’s early. There are told for sequencing conversations but the analytical insights products for conversations are just getting started. Conversation logs can be very helpful due brute forcing where drop off happens, but it’s be great if there were a way to detect unsupported intents from a range of customer conversations, or even told that allowed there brand or service owner to monitor a bot’s performance and then follow up if the bot misbehaved or didn’t understand what the user was trying to do.

I’ve started to play with PullString, and am eager to see what Automat is building.