University students routinely confront fragmented information ecosystems: unstructured documents (e.g., regulations, policies) coexist with structured databases (e.g., individual records), yet most conversational AI systems can handle only one modality and thus fail on complex, hybrid queries that re-quire cross-source synthesis. To close this gap, we propose Academic Query Orchestration, a novel AI-agent framework whose core “brain” is guided by a carefully engineered Super-Structured Prompt to reason, plan, and orchestrate a Specialized Tool Suite comprising a RAG-based Knowledge Retriever and a safety-hardened SQL Executor. The key innovation lies in the agent’s ability to autonomously decompose hybrid questions and fuse heterogeneous evidence into a coherent answer, reinforced by a feedback loop for self-repair. For rigorous validation, we conduct a comprehensive evaluation on established public benchmarks. Our approach achieves competitive performance on single-modality specialized tasks (including complex Text-to-SQL and knowledge-intensive QA) and—crucially—substantially outperforms strong baselines on a challenging benchmark purpose-built for hybrid, multi-source reasoning.