Conversational Storytelling Agent with Satellite Data: Adapting to User Profiles

Although satellite programs like Copernicus now provide abundant Earth-observation (EO) data, non-specialists still struggle to access, process, and interpret it. This paper introduces a swarm-agent framework that closes this cognitive and technical gap. Plain-language questions are converted into structured workflows executed by specialized, modular agents that follow the four classic data-storytelling stages: clarifying the user's question, exploring the data, structuring the narrative, and presenting the results. The system then generates a concise story, supported by visuals and straightforward explanations. A proof of concept (POC) using Sentinel-5P data demonstrates how this approach can track pollutant levels in France's Centre-Val de Loire region. The POC shows how a data-storytelling-inspired swarm-agent architecture can make EO insights accessible, intelligible, and transparent to a broad audience in under 10seconds.