Ultimately, AI agents will be able to perceive (process and interpret data), plan, act (with or without a human in the loop), reflect, learn from feedback and improve over time.