An AI-Assisted Research Automation System for Scholarly Paper Retrieval and Review With Workflow Orchestration

Abstract

The rapid growth of scholarly publications has increased the complexity of conducting rigorous and reproducible literature reviews, while traditional manual review processes are increasingly constrained by information overload, time consumption, and inconsistencies in screening decisions. This study proposes an AI-assisted research automation system that integrates workflow orchestration with literature retrieval and review support mechanisms to address these challenges. The workflow supports multiple stages of the literature review process, including paper retrieval, preliminary screening, and review management, while maintaining human oversight and methodological control. Artificial intelligence, particularly natural language processing through a large language model, is employed as a supportive component to enhance efficiency and consistency rather than to replace the researcher's judgment. Implemented as a workflow-driven process, the system enhances transparency, traceability, and reproducibility in review activities. In the observed execution, the retrieval and preparation stages processed 20 papers in 21.38 seconds, while AI-assisted abstract analysis required 25.02 seconds per paper. These results indicate that workflow orchestration can reduce operational workload and support systematic literature review practices while preserving research integrity and human decision authority.