From Forecasts to Preparedness: Agentic AI for Multi-Hazard Early Warning Services

We developed an agentic AI solution that helps meteorological services prepare early-warning and preparedness reports faster, using operational weather information, local warning rules, and preparedness guidance. Instead of simply generating text, the system coordinates the tasks required to turn multiple sources of meteorological evidence into a structured, traceable draft that forecasters can review and issue.
The solution has moved beyond concept demonstration into cross-country, multi-hazard application. In collaboration with Djibouti’s meteorological service, the system generated a draft preparedness report for the 27 March 2026 rainfall event in less than 15 minutes. Comparison with the official warning showed broad agreement in risk, affected areas, and timing. In Mongolia, the system was applied to a dust event on 18–19 April 2026 and a rainfall event on 29–30 July 2026, with validation from the National Agency for Meteorology and Environmental Monitoring (NAMEM). During Tropical Cyclone Maila in Papua New Guinea on 9–12 April 2026, satellite observations and wind and rainfall forecasts were combined into a preparedness report, and the PNG National Weather Service subsequently recognized its operational value. In Jordan, localization is now underway following a formal request from the national meteorological service for local data integration and Arabic warning services.
These deployments demonstrate that the solution is reusable across countries, languages, and hazard types rather than being a single-purpose AI application. Its modular architecture allows local data sources, warning rules, and preparedness procedures to be adapted while preserving the same operational workflow.
The team is now developing a reusable Early Warning Assistant to support, rather than replace, forecasters. By reducing routine information-processing demands while keeping experts in control, the solution offers meteorological services with limited staffing and technical capacity a practical pathway to strengthen early-warning operations and contribute to the UN Early Warnings for All initiative.