ETD

• agentic AI
• data stewardship
• Intelligent Data Steward
• large language models
• multi-agent systems

This thesis presents the design and development of the Intelligent Data Steward (IDS), a hierarchical multi-agent system that simulates and automates the responsibilities performed by human Data Stewards. In an organizational context, data stewardship is essential for ensuring data quality and effective analytics to support well-informed decision-making. Nonetheless, traditional stewardship practices require significant manual effort, can cause communication delays among stakeholders, and can be prone to human error. To overcome these limitations, the proposed Intelligent Data Steward operates by delegating tasks to autonomous, specialized AI agents, thus increasing efficiency while reducing operational and communication overhead. The proposed system is implemented as a hierarchical multi-agent architecture using the LangChain and LangGraph frameworks. It interacts with external resources to support data discovery, metadata evaluation, data quality assessment, and other stewardship tasks. Additionally, it offers a web-based interface that enables users to interact directly with the system in a natural language, offering a simple and user-friendly interface.