This paper proposes a framework for applying deep reinforcement learning to game AI, aiming to create more believable and human-like characters. It addresses current limitations in deploying machine learning agents in games and identifies key research challenges that could accelerate AI adoption in the video game industry.