Works fine. While I like the idea of two player games (I keep being tempted to do a 2 player network one), I noticed that the sole (I think) 2 player from last year didn't do well because the judges had no one to play against. Adding AI would fix this.
Maybe for each cell, score itself & each surrounding cell - maybe = A*cell_growth + B*own_troops_in_cell + C*enemy_troops_in_cell. Where A, B, C are TBS constants, which could be negative. You may alternatively need weighting inversely proportional to the number of troops in the cell. Maybe B/(own_troops_in_cell+1).
AI for attacking enemy cells could be tricky. I assume you need the total number of your troops piling into the enemy cell, to exceed the enemy troop number. If this is simplified to only considering an attack from one cell, the score for the enemy cell could be zeroed, if you have insufficient troops to take it.
I was thinking, maybe split the troop movements proportionally according to cell scores, but as there are only 5 moves allowed, it seems more likely that a record of the top 5 scoring moves should be kept & those executed. Maybe simplify to only allow one troop movement from each source hex - that would avoid diluting troop effort, which could lead to attempts to capture enemy hexs to fail.
Not perfect, by a long chalk (but then the human player wouldn't have a chance). Don't know if there's room either