Suggested ML technique for recommending user behavior with a GUI-based tool

I have been asked to add the capability to recommend the best way to use a particular GUI program, based on how it has been used by other users before. It is highly targeted to a particular domain, wherein the User uploads some data and then perform a series of different steps one after another and the program can be used to view and analyze the data. Hence, the GUI program ends up being used in different ways by different Users. We would like to build a recommender system to learn the way in which it is used most efficiently (for certain types of input data) and then make recommendations to the user. What ML technique (along with a mature toolset) would work out? I am new to this, so looking around it seems that HMM might work out ? I found some relevant discussion here - http://stackoverflow.com/questions/6990230/supervised-learning-for-user-behavior-over-time