The presentation will focus on the coalescence of technology that is enabling geospatial AI use cases to take place across the enterprise. It will feature a playbook for pivoting away from the legacy GIS operating model (teams and technology) into a modern AI + Cloud model.

The Census Bureau will showcase its use of GeoAI to automate nationwide change detection and to extract building footprints and road features to enhance the Master Address File/Topologically Integrated Geographic Encoding and Referencing System (MAF/TIGER) System to support a more complete and accurate decennial census. By automating these workflows, the Census Bureau can devote more resources to complex geospatial challenges in preparation for the 2030 Census.

For modern labor organizing, spatial data is a critical multiplier. At Working America, its true power lies in its ability to driving cultural transformation by dismantling the gatekeeper model of data analytics. By building an ecosystem that equips staff to answer strategic questions directly, we have shifted the paradigm of data access. This talk maps our journey, exploring how we are navigating engineering hurdles, governance, and user adoption while charting our next frontier in spatial analytics.

Most geospatial embeddings and location analytics rely on static socioeconomic survey estimates, commercial point-of-interest catalogs, and fragmented foot-traffic samples—capturing what a place looked like in the past while missing the dynamic human intent, economic momentum, and behavioral shifts that drive real-world change. Google’s Population Dynamics Insights (PDI) pioneered a new paradigm in geospatial AI: rich privacy-preserving 330-dimensional foundation embeddings (“fingerprints”) learned by fusing direct physical-world observations from Google Maps points of interest, pedestrian mobility patterns, and environmental data with Google’s proprietary Search trends. By encoding local search intent alongside physical mobility and built-environment structure, PDI embeddings capture multi-dimensional human dynamics and leading behavioral indicators that static structural models and conventional location providers cannot replicate. Drawing from two years of real-world deployments across epidemiology, population modeling, insurance risk, and retail demand forecasting, this talk highlights the latest developments in geospatial foundation models from research to production. We detail PDI’s expansion to fine-grained, global S2 Level 12 cell-based embeddings with continuous monthly temporal snapshots, enabling longitudinal change detection and cross-border generalization without manual feature engineering.

Population Dynamics Insights is part of Google’s broader Earth AI and geospatial analytics portfolio. Our geospatial analytics datasets and models, integrated with BigQuery and Google Cloud, enable practitioners to combine PDI’s high-dimensional predictive embeddings with interpretable physical-world datasets such as Places Insights and Environmental Insights, alongside agentic geospatial workflows to achieve both state-of-the-art predictive accuracy and transparent, auditable spatial decision-making at scale.

Traditional economic forecasting at the city level can mask significant variation in economic outcomes across areas within the same city. Oxford Economics has developed a novel hyperlocal forecasting framework for the US, which disaggregates cities into uniform hexagonal grids and models economic activity using a quantitative spatial model. Using case studies from New York, Los Angeles, and elsewhere, we demonstrate how hyperlocal forecasts reveal within-city divergences that are invisible at the city level. These insights support more targeted decision-making for investment, site selection, and policy at the scale where those decisions are actually made.

Large language models have transformed AI by learning general-purpose representations of text. A similar shift is beginning in geospatial AI, where foundation models learn reusable representations of place instead of engineering new features for every task. This session introduces PlacePulse, which represents every location in the United States as a compact embedding learned from rich structural data. These embeddings act as a reusable fingerprint of place, supporting diverse spatial analysis workflows. Through real-world examples, we'll show how geospatial embeddings enable applications from similarity search to predictive modelling, and discuss the opportunities and challenges of bringing foundation models into production.

How an investment firm could use Cortex Code and Snowflake to build a smarter entry strategy for utility-scale solar, starting with the land.

For years, spatial analysts and field operations battled rigid queries, manual updates, and high-maintenance custom applications. It’s time for a shift. In this session, we will explore how combining CARTO with AI Agents completely modernizes this legacy workflow. Discover how we transformed static analytics into dynamic, conversational spatial applications. We will dive into how CARTO can speed up time to delivery, explain how CARTO’s AI agents orchestrate and automate spatial analysis, and share the major leaps that can be achieved in operational efficiency.

Climate events are occurring daily across the globe, and without significant emissions reductions, their intensity and frequency are expected to grow. ICE Climate provides insights into the material impact and financial losses from current events, alongside tools designed to project future climate related risk. These models are built on high resolution, geospatial data and analysis. This presentation shares real-world examples using data and visualization to explore how climate risk is unfolding today and how it can be assessed into the future.

A good audience-movement visualization looks effortless — a smooth glide across the city. Underneath, it’s a high volume of data with unique challenges: signal sparsity, trajectory inference, snapping points to road networks, and reconciling voluminous but rarely cooperative device data. This session looks beneath the waterline at how we turn noisy pings into OOH decisions stakeholders trust, and how to visualize calm waters while the underlying data is far more sophisticated.

AI is becoming a powerful interface to enterprise data, but intelligence without context has limits. Business decisions happen in the physical world: customers live somewhere, assets operate somewhere, events occur somewhere, and risks unfold across space and time. Spatial intelligence gives AI a sense of place. Vector search helps it understand semantic similarity. Enterprise data provides authoritative facts. And agent memory gives it something increasingly important: a history of what it has learned, what decisions have been made, and what happened as a result.

In this keynote, Oracle explores how these capabilities can converge to create a new generation of context-aware AI. Instead of simply asking where something is, organizations can ask what is happening there, why it matters, how it compares to similar places, and what previous experience suggests they should do next. As AI agents become more capable, spatial reasoning and memory can move beyond specialist GIS workflows and become foundational capabilities for enterprise applications.

The future of spatial intelligence is not simply AI with a map. It is AI that understands place, context, and history — and uses them to make better decisions.

Geotab connects over 6 million commercial vehicles worldwide, generating a continuous stream of telematics data. Turning that data into geospatial decisions — at scale, for fleet managers everywhere — is exactly where Carto comes in.

Model Context Protocol (MCP) is changing how enterprise platforms connect to AI — and to each other. In our first experiments linking Geotab's telematics data to Carto's geospatial analytics via MCP, we were able to go from live vehicle locations to spatial enrichment queries (nearest POIs, fleet clustering, status mapping) in minutes, using natural language and with PII controls enforced throughout.

The combination opens up powerful new possibilities for fleet managers: richer location intelligence, faster insights, and geospatial