We said an intelligent agent has always a objective (goal). Which is the objective of an agent which is able to understand and formulate phrases in natural language? As an example it could be to entertain someone in a waiting room (possibly without that this someone notices he is speaking with something much more similar to a computer than to a human). If we want to be a little more professional we could say that the scope of the agent could be " simply " exchanging useful information in order to catch up some other objective. In this case the ability of communicating using natural language would have to be considered like an additional action the agent can use to catch up its objective. Now suppose we need to build an agent who is able to move in a building (imagine something like a secretary robot). One of the problems we have to solve would be that one to supply to the agent the map of the building (in digital format obviously). Suppose now we need to move the agent in another building. It would be necessary to change the agent's map. Well, why not to give to the agent the possibility to ask information about his position, about the position of the rooms he needs to reach and all what else he needs to auto-generate his own map, according to the received information? The answer is simple: actually it is much more easier and convenient to create or modernize a digital map, rather than studying the problematic connected to parsing and understanding natural language. However these problematics have the fascination of the (perhaps) irresoluble problem. That's the basic reason for which we are going to talk about natural language.
Fig.1.1: intelligent agent.