Earth observation use cases with AI

This workshop introduces the concept of agentic EO systems, where AI agents combine planning, tool use, and reasoning to autonomously perform multi-step geospatial analysis from natural-language requests. Drawing on recent developments, the session explores open-source architectures, data sources, and AI frameworks that could support similar capabilities in research and public-sector contexts.

Participants will gain a practical understanding of how emerging agentic AI approaches can be applied to Earth observation (EO) workflows.

Benefits for the attendees, what will they learn:

• Understand the key characteristics of agentic AI systems, including planning, tool use, memory, and multi-step reasoning in an EO context.
• Learn how natural-language interfaces can support complex geospatial analyses without requiring extensive remote-sensing expertise.
• Become familiar with open-source and open-data components that can be combined to build agentic EO workflows, including Sentinel and Landsat data, EO APIs, open-weight models, and agent frameworks.
• Understand the architecture of AI agents interacting with EO data, from data discovery and processing to analysis and reporting.
• Explore practical EO use cases such as environmental monitoring, disaster assessment, urban expansion analysis, and infrastructure detection.
• Develop a critical perspective on the opportunities, limitations, and risks of agentic EO systems, including issues of reliability, explainability, transparency, evaluation, and governance.
• Exchange ideas with peers and identify potential applications of agentic EO approaches within their own research, projects, or organizational contexts.

Level

Intermediate. Participants should have a basic familiarity with Earth observation, geospatial data, or machine learning concepts. No advanced programming skills or prior experience with agentic AI systems are required.