There is a tendency to think that making a product conversational automatically makes it more useful. I am not convinced that is true. A chat interface can make a product feel modern, but that is not the same thing as solving a meaningful problem.
I find it more useful to start with the workflow. What is the user trying to get done? Where is the friction? Is there something that currently takes too many steps, requires too much searching, or depends on expertise that the user does not have? AI becomes interesting when it can remove some of that friction rather than simply put a natural language layer on top of an existing workflow.
This also changes how I think about measuring AI products. Traditional metrics like time spent or frequency of use can be misleading. Sometimes the best AI experience is the one that lets a user finish the task faster and leave.
The important question is whether the product creates useful outcomes, not whether it creates more engagement for its own sake.