I generally agree with your point that forcing users into new and unwanted habits via a bot in a one-off conversational UI is troublesome (it sounds like you identified this as the trending pitfall for bot-makers). I also agree that better apps, and not bots, will replace apps (per Dan Grover)
However, I do not think all useful bots have to replace humans — this view shortchanges the power of a bot. Some bots will augment the ability of humans to complete work, serving as powerful tools for humans to leverage in traditional activities, rather than replacing humans themselves. In other words, bots can take on work that was previously impossible or unfathomable for normal humans. Not every useful bot will replace human activity if the bots are helping humans take on new work and greater challenges. Do you disagree, or am I misinterpreting your take?