The next step could be to develop a method that predicts emotions and feelings, but I am not sure that this will always be true. It could also be the case that AI can only be developed to a certain point, because when emotions and intuition come into play, it may not be possible to reach a real conclusion.
If the level of AI really does rise to this level, we cannot know whether it will be good or bad. Much can be said about the emotional development of AI, but until that happens, it is difficult to say for sure where things will go or whether it will not happen.
It's possible to develop an AI which identifies emotions through micro facial expressions. It's not 100% accurate, but it can work with a decent error margin. There are emotions considered primary ones which are easily spotted, such as: anger, joy, sadness, happiness, fear and disgust.
Of course the market will take advantage of this possibility to identify emotions on people, and then offer them products and services accordingly to what they seem to need in the present moment.
On the other hand, I don't think an AI can replace a therapist, for an example. It must feel really strange to talk to a soulless being and the effects of the treatment must be incomplete in the end, as it's mainly the connection between two human beings which result in progress along the sessions.