This paper introduces a groundbreaking approach to Information Retrieval (IR) centered on agentic AI, which leverages intelligent, adaptive systems to enhance document retrieval. By combining graph theory for structural alignment with the BERT model for semantic analysis, the method evaluates document relevance holistically, considering both data structure and textual meaning. This dual-layered approach ensures a deeper understanding of documents beyond traditional techniques. A distinctive feature of this system is its adaptive AI framework, which dynamically adjusts the balance between structural and textual relevance based on user interactions and feedback. This adaptive capability empowers the system to continually improve its results, tailoring search outputs to meet evolving user demands. Importantly, it fosters collaboration by aligning results with the strategic and problem-solving needs of users. The research specifically targets XML documents, a critical format widely used in industries like finance, healthcare, and logistics for structured data exchange and representation. XML’s importance lies in its ability to standardize information while accommodating complex hierarchies, making it indispensable for large-scale data systems. Evaluations on a large multimedia dataset show significant gains in retrieval accuracy and contextual relevance, demonstrating the potential of agentic AI to handle the unique challenges posed by XML-based searches. By integrating advanced AI with XML’s industrial relevance, this approach highlights the transformative potential of adaptive, intelligent IR systems to support collaborative workflows, strategic decision-making, and improved information management in complex environments.