1. As a customizable opponent. An AI can be programed to go easy on the player and help them get accustomed to VizionEck. An AI can also be programed to emphasize certain tactics and spar with the player for training.
2. As a stand in for human players when no local or online friends are avaliable.
Reason two is especially hard to achieve for VizionEck since the game is focused on strategy. An AI with perfect motor skills but poor planning would be easy for a human to beat. As such, my current approach is to model the AI off turn based strategy games.
Every turn (1 frame) the AI starts fresh and re-calculates what it should do. This vastly simplifies things because only one general AI function is needed. I'm still working on structuring this function though. The standard technique is to have a list of possible actions and calculate the desire to do each action, and then only do the most desired action. The problem with this method is that it works best when the actions are clearly defined and have little in common. In VizionEck, there's significant overlap between actions and the AI needs to have the ability to find a "common ground" between things if it makes sense.
I'm still in the very early stages of the AI. Hopefully next week I can come back and report some significant progress.