Although bots have been around for a long time, most of them have required precise commands. You're right that imprecision is why it can be successful, and the key to making this work is having manager interfaces like Siri, Cortana, Alexa, or Hound which can add layers of abstraction over simpler single-purpose bots. As we standardize communication protocols between bots, it will be easy to send a header describing abilities and responses so that the user interface can both parse the capabilities for the user and send feedback about how to be more responsive to the bot and its programmers. In the end, if this works, we should have a massive amount of machine learning taking place as bots improve each other.