Despite a decade of promises around educational predictive analytics, from identifying at-risk students to providing personalised feedback, the practical impact of such systems remains marginal in many institutions. Drawing on experience as both an IT professional and academic working with enterprise educational platforms, this paper examines whether Large Language Models (LLMs) will meaningfully transform this landscape or merely reframe existing challenges.

Rather than focusing on theoretical capabilities, this analysis considers what AI can realistically deliver within the constraints of complex institutional systems, limited educator time, and competing pedagogical priorities. Even sophisticated AI provides minimal value when it either confirms what educators already know or identifies problems that cannot be practically addressed.

The current moment presents unprecedented opportunities to use powerful analytical and automation tools. However, the "not-yetness" of these initiatives often stems from unclear objectives and implementation barriers. This paper outlines common challenges educational institutions face when adopting AI, identifies necessary practical steps for successful implementation, and establishes realistic expectations for impact. By bridging technological possibilities with institutional realities, we can move beyond hype to develop AI applications that genuinely enhance educational outcomes in sustainable, meaningful ways.