A big part of playcalling is countering what the opponent is doing specifically against your scheme in the current game. If you had a system that can analyze film in real time to identify defensive schemes, mismatches, player positioning, players who are more likely to make mistakes in the current game, etc, you might have a shot. But if you can't do that, then you can't call Zach Wilson's QB draw against USC that went for a TD, for example.

Also, AI needs lots and lots of data in order to learn how to be good at something, and I strongly suspect that the length of a game is not long enough to collect enough data to be consistently better at playcalling against a specific team on a specific day than an experienced coach would be.