One of the most effective ways to "reduce" the discourse directions is to turn the Chatbot from passive to active. In other words, the Chatbot is the one who asks the questions in the conversation, and an end user supplies the answers. This creates a situation where the conversation is defined and confined to the subjects that the Chatbot is designed to handle, while everything else will be answered something like "Sorry, I did not understand ..."
Limiting the conversation greatly improves the ability of the Chatbot to respond, since now the information coming to the Chatbot from the user is more concise and known in advance. In many cases, the Chatbot limits the answer to only "yes" or "no".
But... that's not convenient for the users.
If there is a conversation where one side "interrogates", the other side will naturally feel uncomfortable. Even if everything being asked is essential for the situation, the interrogee will probably still feel uncomfortable.
It helps developers - not necessarily the organization
The large beneficiaries of the restricted Chatbot method are actually the Chatbot developers. By limiting the discourse, Chatbot development becomes easier and faster. The developer’s need, however, does not always meet the interests of the Chatbot owners, and this point must be taken into consideration when approaching Chatbot.
Conversation does not flow like this
One of the conditions for a pleasant conversation between two participants is that the conversation is "rolling". A “rolling” conversation does not have a permanent and predictable structure. It is possible to start with a specific subject, and from there to move onto another topic, and perhaps at the end you reach the question that should have been asked first.
Let's look at three questions in car rental for example:
“When would you like to rent the car”?
“What kind of vehicle would you like to rent?”
“Where would you like to rent the car from?”
It is reasonable to assume that the client will want to clarify these details in this order. But what happens if a customer does not know how to drive cars with manual gear? Perhaps the first question, and the most important one for him, would be:
“Do you have automatic gear vehicles?”
Should such an approach be considered? If you want the Chatbot to respond in the same way the sales representative would, the answer is: for sure.
But when the Chatbot is built with a narrow conversation narrative, there is no way to start a conversation from the middle and then go back to the correct "up hill" position in the decision tree. In practice, for such a client, the simple tree starcatcher Chatbot is inefficient and can not provide a solution.
The question you need to ask yourself is "how many of my clients are customers with special requirements?" The answer to this question is not the same for every business or organization, but in general, it can be said: Human beings are special, and have special requirements. Each one has his own unique way.