The advent of Large Language Models (LLMs) has fundamentally disrupted our thinking about educational technology. Their ability to engage in natural dialogue, provide contextually relevant responses, and adapt to learner needs has led many to envision them as powerful tools for personalized learning. This emergence raises important questions about their relationship with Intelligent Tutoring Systems (ITS), which have long been the gold standard for computer-based personalized instruction through their structured, discipline-focused approach. While the potential for integrating these technologies is compelling, significant theoretical and practical challenges remain. This paper examines these challenges and proposes new ways of conceptualizing the relationship between LLMs and ITS to enhance both personalization and learning outcomes. [This chapter was published in: "Design Recommendations for Intelligent Tutoring Systems: Volume 12 - Generative AI in Intelligent Tutoring Systems, Army Combat Capabilities Development Command - Soldier Center, Orlando, FL," 2025.]