Recommender systems are ubiquitous in today’s digital online platforms, powering personalized experiences on platforms ranging from e-commerce giants like Amazon to media services such as Netflix. These systems utilize algorithms to recommend items tailored to individual user preferences and historical behaviour. Recently, the emergence of the Agentic AI paradigm, defined by goal-oriented behavior, contextual awareness, and minimal reliance on human instruction, offers a transformative path forward. This paper provides an overview of recommender system models, ranging from foundational algorithms to the latest innovations driven by Large Language Models (LLMs) and multi-agent frameworks. We explore how Agentic AI is reshaping the design of next-generation recommender systems by enabling dynamic interaction, real-time learning, and simulation-based personalization. Through a comparative analysis of current models, we highlight the benefits of the paradigm—such as improved adaptability and autonomy—alongside persistent challenges, including model interpretability and system complexity. By synthesizing current literature and identifying key trends, this paper aims to offer a roadmap for future research at the intersection of Agentic AI and recommender technologies.