I am very happy with your decision to improve graphics and switch to 3d. I would like to suggest that you start collecting player data.
If the game sells as well as we all hope for there should millions of player turns with even more different “battle situations”. If there is a data base with all these situations modders or game designers can just choose the most adequate of these battle situations. For example: most damage done by a human player in this situation in one round of play or most damage done by a human player in two rounds of play. You could also choose least damage taken by a human player in this situation in one round of play or movement done in this situation by the most successful player on this map (or the overall most successful player). This approach might also work if it is only almost the same “battle situation”. For instance if the same outcome can be reached. It would be just great for modders to experiment with this kind of data.
Sure this data base would just get useful relatively late in the product life cycle of the game but it might be great for advertising. You could just ask players after installing the game if they would like to contribute with their anonymous game data to an experiment with a “learning” AI. It is learning because the more human turn data there is the better the chance of getting a matching situation solved by a competent human player. I hope this does not involve to much of an investment in terms of time and money for you. I think that a challenging AI is the most important feature for this kind of game, maybe just as important as game mechanics or looks. It would be great if I could get a comment on this page whether this idea is any good. Thank you very much for your time reading this.