Harmonizing Subsurface Data Platforms for Agentic AI in Upstream: A Dual Layer Architecture Using OSDU and Microsoft Fabric

Even in the present day and age, the upstream oil and gas industry continues to face significant challenges in integrating siloed subsurface data across domains such as seismic, wells, reservoirs and production. As part of ENERGYai – an industrial AI initiative advancing upstream intelligence – we present a dual-layer data architecture that harmonizes two platforms: the Open Subsurface Data Universe (OSDU) as a canonical subsurface data layer, and Microsoft Fabric as a semantic and analytics layer. This integrated architecture creates a robust foundation for agentic AI in oil & gas upstream workflows. We outline how diverse data types (seismic surveys, well logs, petrophysical measurements, production records, etc.) are unified in OSDU, then virtualized into Fabric's Lakehouse environment via OneLake. This approach retains a single source of truth while enabling rich semantic models and cross-domain analytics without duplicating data. We also describe specific AI agents (for metadata tagging, document parsing, QA/QC, and ingestion planning) that capitalize on this architecture to automate data management tasks. Key outcomes include improved data quality and lineage, faster data ingestion throughput, and enhanced user access to data. Ultimately, geoscientists and engineers are empowered to spend less time on data wrangling and more on value-added analysis, demonstrating clear efficiency of gains and business value.