Esri demonstrates AI extraction of boring logs into ArcGIS map data

Esri has documented a customizable workflow that connects external language models to standard ArcGIS Pro text analysis tools. It demonstrates extracting PDF boring-log contents into structured soil-layer records for mapping, rather than manually transcribing those fields.

For engineers assessing ground conditions, this provides a route to make archived subsurface information accessible alongside mapped boreholes. Its operational value depends on extraction quality and whether project data can be sent to a third-party model provider.

Esri’s walkthrough shows how geotechnical boring logs can move from PDF files into structured spatial records within ArcGIS Pro. A custom deep learning package connects the GeoAI text analysis framework to a web-hosted language model, extracts requested fields, and returns one row per soil layer. The demonstrated map exposes classifications, depths, field notes, and refusal information through borehole pop-ups. Developers can adapt the extraction prompt and output schema for other report types, while users can select a provider and configure model arguments without opening the package. The significance is a reusable route from document archives to GIS attributes, rather than a standalone document-reading tool. Implementation still requires an Advanced license and attention to data governance: processed material is sent to the external provider. The article demonstrates the workflow but supplies no extraction-accuracy benchmarks or measured time savings.

Evidence supplied by Esri.