Training the BOT

I will first start by saying that I love the BOT.

With that said, you (FreshDesk)need to have some work on it to improve it.

• Asking questions:
  Currently the BOT does not ask any questions when it cannot have an answer - it still tries to pull a solution from the KB and provides poor service.
  
  For a statement by the user saying "it does not work" - the BOT should learn to ask a followup question (that we can provide and map), asking things like - 'Did you receive any kind of error?', 'Please elaborate on the issue' or simply ask the user to rephrase the question.
  
  
• Answers not through articles:
  The bot doesn't know how to cope with basic questions that users do tend to ask BOTs, like - what is your name? are you real? things that are related to the BOT and should not reside in an article.
  
  
• Selecting a specific answer: (Agent side)
  When you set up the BOT - you go on asking it many questions to see how it answers, mapping the relevant answers when the BOT fails to do so on its own.
  With that said - you completely ignore the mapping - still giving the user three options to choose from. What is the mapping for if you ignore it?
  Why not return the answer I just mapped?
  
  
• Selecting a specific answer: (User side)
  When I ask a user as a user - I get three optional results.
  After I clicked on one - i cannot mark it as the correct answer.
  Instead - I need to go back after reading it and tell the bot that his 3 answers were helpful.
  
  How will the BOT learn which of the three was correct?