Agentic Recommender Systems

Recommender systems are entering a new era. As large language models and agent-based AI reshape what's possible, recommendation is no longer just a prediction problem: it's becoming a collaborative process where intelligent agents reason, communicate, coordinate, and adapt to ever-changing user needs.

Agentic Recommender Systems is a field that is exciting but still fragmented. Spanning recommender systems, LLMs, multi-agent systems, software engineering, and human-computer interaction, this book offers researchers, students, and practitioners a much-needed shared vocabulary and conceptual foundation. It moves from first principles to the frontier: grounding readers in how agentic approaches differ from traditional recommendation pipelines, proposing a general framework for reasoning about multi-agent recommendation, and tackling the open challenge of evaluation: datasets, metrics, baselines, and the limits of current methods. Along the way, real-world applications, architectural patterns, and the fast-growing ecosystem of agent frameworks and orchestration tools bring theory into practical focus.

Rather than claiming to have the final word on a still-maturing field, this book is designed as a resource to help readers navigate preference elicitation, continual user modeling, multi-agent coordination, and the broader implications of increasingly autonomous recommendation systems. Whether you're a graduate student entering the field, a researcher pushing its boundaries, or a practitioner building the next generation of intelligent recommendation technology, this book provides the coherent, comprehensive starting point the field has been missing.