Anna

Anna is a governed execution layer for AI-powered analysis from Geo Orchestration AI. It addresses a problem the provider describes in model-led AI systems: when the same language model interprets a request, decides what to do, triggers tools and explains the result, uncertainty in language becomes uncertainty in execution.

Anna separates the work into three roles. Language models interpret a plain-language question, select among registered datasets, metrics, entities and workflows, and explain the completed result. Deterministic code performs the data access, joins, aggregations, geospatial operations (overlays, administrative rollups, location-based analysis), pipeline calls and output generation. Provenance is captured during execution, so a user can see the data used, the filters applied and the workflow steps followed. Outputs include maps, charts, tables, dashboards, downloadable files and narrative summaries.

The provider calls Anna domain-agnostic by architecture, with an initial focus on health, geospatial intelligence and social data intelligence. Published examples include drought health risk by ZIP Code in Texas and average distance to a women's health clinic by ZIP Code in Pennsylvania. Developers can bring their own models, data pipelines and workflows through API-based integration. Registered data can include satellite-derived layers: the provider lists its Poseidon Plastic Tracking System among the solutions powered by Anna, and Poseidon maps coastal plastic from multispectral satellite imagery (Sentinel-2 in its 2023 Arctic pilot). Announced pilots include WaldoAir (aerial imagery combined with health risk data for disaster response, 2024) and Skymantics (DataGenesis synthetic population data, 2026). In August 2026 the provider reported a 100x reduction in language-model token use compared with agentic approaches, a figure it describes as 'tested via Perplexity'.