How do they surface insights without being programmed to parse / understand what an insight is?
Why would Christina assume to know what a spike is. Perhaps I already know WHY it spiked (I sent an email) and I need to parse specific UTM or other references to segment the outreach. Is Christina going to offer a dozen options for me?
The problem with the concept of “conversational UI” (not a phrase I expect to become canonized historically) is that you’re being talked to first in all of these examples. Those are merely guesses.
The audience is the one who starts the conversation, which is essentially experience map / funneling (you know, the buttons on the left of analytics). The reports and segments that I use with analytics are vastly more complex than “hey would you like to know traffic stats hurr durrrrr.”
Rather than having a conversation about the data, I set up automated reporting based on that data and the reports generate clear datapoints that a brained individual, not AI, can determine.
EG, here is the timeline and this spike happened when the email went out which was 25% of total traffic.
I don’t need to play guessing games with a chat bot to determine that.
I mean, if that was the case, why not just have the chat bot SHOW you the data + conclusion? That’s all that the upper management wants, if they are to implicitly trust the coding that begins the conversation why not trust the data processing at the other end?
It’s like saying “in the future your refrigerator will ask you a question regarding what it assumes you want to eat”
No thanks, I’ll just reach in and find it. OR, if you’re going to prompt me with suggestions and I trust you, just dispense it.