Good point, but in this example the user is mute and the interface is omniscient, “would you like to find out why”. One of the best things about CUI is the ability to extract value from your natural language responses. Part of that comes from assuming the user has knowledge you don’t and prompting them to share it.
Applied to your example that might look like this:
Christina: looks like there’s an unusual spike in traffic, do you know why that is happening?
Lakshmi: We just had a big PR event.
Christina: If you have more of these I can compare success rates. When is your next event?
It’s not even necessary in this case for the system to know what a “PR event” is to provide the value of comparing them. Also the more users mention “PR event” the more you know that your product’s knowledge base needs to be extended in that area, something pure web analytics would never uncover.