From Metadata Strategy to Practical Data Infrastructure

This summit takes place from September 6 – 7, 2026 in Bolzano, Italy. It focuses on the implementation of metadata standards — what works today, what needs to evolve, and how organizations can apply existing and emerging approaches to real-world challenges. Dr. Benedikt Gräler will introduce a multi-agent AI framework designed to reduce friction across the entire metadata lifecycle.

“From Tedious to Automated: A Multi-Agent AI Approach to Reducing the Metadata Curation Burden”  (Dr. Thore Fechner, con terra GmbH, Dr. Benedikt Gräler, 52°North GmbH)

Abstract: As Spatial Data Infrastructures (SDIs) evolve, the bottleneck for effective data reuse has shifted from technical accessibility to semantic clarity, as traditional manual metadata management often results in incomplete documentation that hinders modern AI agents. In this presentation, con terra and 52°North, working under the Spatial AI Nexus (SAIN) Münster, introduce a multi-agent AI framework designed to reduce friction across the entire metadata lifecycle. By leveraging Retrieval-Augmented Generation (RAG) and schema-aware reasoning, the system employs three specialized agents: a Textual Analyst for semantic integrity, a Structural Validator for schema compliance, and a Field Expert that bridges raw data with policy by analyzing both linked data and the raw datasets themselves to suggest metadata improvements. This approach automates the generation and refinement of metadata to ensure datasets are AI-ready and reliable for autonomous GIS workflows. Furthermore, this framework extends to a second project context, where LLM-based natural language support enables intuitive harvesting and semantic searching of metadata, significantly enhancing the efficiency of geospatial data discovery.