I have the following problem. There is a software that I've written some time ago. Users enter customer's data in the system and there is a limited number of things (actions) that they can do/add - connected directly to that customer (for example: customer called, customer sent email, I sent document to the customer etc.). There are about 10 actions that can be done. Now I have thousands of historical data (sequences of actions per customer). I would like to add prediction for the next best/most likely action (or actions) when user opens the customer's window. Something like assistant saying "perhaps you want to..." and gives 1-2-3 most likely actions that user would do with this customer. So far I have found only one way to achieve something like that, and it's by using Markov Models (treating a list of each customer's actions as sentence and predicting next "word"=action) but I would like to use some more advanced method. I was thinking about RNN/LSTM but I cannot really find any examples or similar problems on the web. Perhaps this problem has another name and I am just searching for the wrong stuff. Any help will be greatly appreciated.
What would be the best approach? I would prefer examples in C# but any language can be considered.