More than fifteen thousand people filled the San Diego Convention Center for Esri’s 46th annual User Conference this year, with roughly that many more watching by livestream from 117 countries. Taken session by session, the program is a wide survey of a mapping platform still expanding in every direction. Taken together, it resolves into a smaller number of recurring bets — about where AI belongs in a GIS, about how far one platform can scale, and about what all of it is ultimately for.
More than 15,000 people filled the San Diego Convention Center for Esri’s 46th annual User Conference, with an equivalent number watching the livestream from 117 countries, according to Esri founder and president Jack Dangermond.
“The purpose of it is fundamentally all about helping you and helping your work and advancing this community that we live in,” Dangermond told the room. Across two days of product announcements, enterprise case studies, and awards, the platform was framed less as a product line than as infrastructure for work its users were already doing — analysis, response, planning, and record-keeping.
The Community Snapshot, Read Forward
Dangermond opened with a curated selection of user-submitted maps organized loosely by theme. Environmental and conservation examples included recovery monitoring in the Amazon Basin, wetlands and forest mapping in Florida, Sylvia Earle’s global Hope Spots marine protected areas map, and work led by Tompkins Conservation.
Natural-resource examples ranged from geologic mapping and mineral claims in British Columbia to subsurface geophysics, crop-health monitoring in Australia, vineyard management in Northern California, and sustainable cocoa production in Cameroon. Land-records and cadastral work included Washington, D.C.’s modernization effort, Sweden’s nationwide parcel fabric, and 3D cadastral mapping in Istanbul. Planning examples spanned Los Angeles urban design, low-carbon village planning in Indonesia, and a global land-use project modeling carbon emissions and sequestration to 2050.
Other categories included indoor and campus GIS; transportation and asset management; utility and telecommunications networks; public safety and disaster response; demographic and statistical work; and private-sector applications from logistics to economic development. The breadth mattered because many of those subjects returned later as full case studies rather than passing references.
Dangermond closed the segment by recognizing this year’s Special Achievement in GIS Award recipients. The morning survey was, in effect, a preview of the afternoon’s actual subject matter.
The afternoon technical program opened with enterprise GIS deployments at San Francisco International Airport, where a mapped digital twin supports an integrated operations center, and at a new Vietnamese seaport, where a geospatial twin combines terrain, underwater, ecosystem, ship, cargo, and equipment data.
Esri architect Sam Libby organized the technical program around four themes: artificial intelligence, deep mapping and analytics, enterprise systems, and outcome-driven digital twins. Those categories also structure what follows, rather than the order in which sessions actually ran.
AI Kept Inside the Map
The plenary’s technical segment covered Esri’s expanding platform, including new agentic mapping web components for JavaScript aimed at increasing developer productivity and opening ArcGIS to AI-assisted development workflows, alongside game-engine-based visualization capabilities.
Three focus products were highlighted specifically: ArcGIS Business Analyst for location-based market and community analytics, used across government, commercial, nonprofit, and community organizations for site selection and service area optimization; ArcGIS Urban for urban design and planning; and ArcGIS AllSource, Esri’s intelligence analysis and decision-support platform built for analysts working across defense, law enforcement, and civilian intelligence workflows.
Ismael Chivite, Esri’s product manager for geospatial AI, walked through what he described as a unique category of artificial intelligence built specifically to solve geospatial problems. AI assistants embedded across ArcGIS are included with user types at no additional cost, are already available in ArcGIS Online, and are coming to ArcGIS Enterprise before year’s end. Live demonstrations showed a natural-language assistant in ArcGIS Pro building charts, running hotspot analysis, and adjusting symbology on New York City Citi Bike origin-destination data; a scripting assistant in ArcGIS Notebooks writing and explaining Python code to fetch bike-share station data and run a hotspot analysis; and a new Item Details assistant that drafts layer summaries, descriptions, and field metadata automatically.
Esri also previewed a new foundation model, the Geospatial Vision-Language Model, built in partnership with Amazon Web Services using large volumes of global geospatial imagery. The model, referred to as GeoVLM, combines image understanding with natural-language prompting and can perform multiple deep-learning tasks, including pixel classification, object detection, counting, and image description, from a single model. A demonstration showed GeoVLM detecting trees, buildings, and roads from drone imagery in a coastal village in Belize, then applying the same prompts to World Imagery basemap data in Las Vegas and to crop classification tasks elsewhere.
Esri also introduced agentic mapping apps, a new category of conversational application built on top of a web map using the Data Explorer instant app template or developer SDK. A demonstration built around a Hawaii logistics and preparedness scenario showed a chat interface filtering fuel depot facilities by supply level, cross-referencing an affected facility’s district demographics, and declining off-topic questions to stay within the authoritative scope of the underlying map data.
A separate demonstration showed ArcGIS content connected to Microsoft Copilot Studio through the Model Context Protocol, allowing a custom agent to query a planning department’s proposed-project layer, identify intersecting flood hazard areas and federally protected lands through a geoprocessing model, and return a permit-status report with a linked map for verification.
That same instinct — AI confined to a specific, authoritative dataset rather than open-ended conversation — showed up again later in the day, on the ground rather than in a product demo. In Allentown, Pennsylvania, the city’s two-person GIS team demonstrated a new AI assistant for ArcGIS Solutions capable of generating a working application, in this case a public art inventory, directly from a natural-language prompt — the same generate-from-prompt pattern as the Item Details assistant, applied to an entire app rather than a single field. At NextEra Energy, an internally built tool called Boundary Hunter uses Claude to extract site-boundary information from documents and generate map polygons, a workflow the company said had reduced tasks that previously took hours or days down to minutes — the model doing the extraction, a person still owning the resulting boundary. And the National Weather Service’s in-development Ask NWS tool, discussed in more detail below, follows the identical logic in a higher-stakes setting: it translates authoritative weather data into plain-language public safety messaging, but a forecaster reviews and verifies every message before release.
Across all four of these — the agentic mapping app, the Copilot Studio integration, Boundary Hunter, and Ask NWS — the AI is doing real work, but none of it is a general-purpose chatbot. Each is scoped to a specific authoritative dataset, and each keeps a human or a defined data boundary between the model’s output and the decision that follows it. That is a more specific and more defensible claim about Esri’s AI direction this year than any single demo makes on its own.
New Rigor in Analysis, and One Tool With Two Uses
Lauren Bennett, Esri’s product engineering director for spatial analysis and data science, and colleague Alberto Nieto introduced three new analytical tools aimed at bringing more rigor to common mapping decisions. Evaluate Bin Sizes for Point Aggregation tests multiple bin sizes against a dataset, in this case a map of over a million GPS points of Mediterranean Sea vessel traffic, and scores each for how well it balances summarization against preservation of spatial patterns, replacing a process that previously relied on visual intuition. Volume Percentile Contours, demonstrated against 30 years of U.S. flash flood data, creates percentile-based zones of concentrated point data that can be compared across multiple hazard types to identify areas of overlapping risk.
A third demonstration used new file knowledge graph capabilities in ArcGIS Pro, built from OpenStreetMap road network data, to run centrality analysis identifying which bridges and overpasses are most critical to a transportation network’s overall connectivity, surfacing some less-obvious chokepoints alongside well-known ones such as Washington, D.C.’s Key Bridge and Memorial Bridge.
Peter Erlenbach demonstrated recent advancements to ArcGIS Pro’s land records and parcel management tools using a live editing example from Sheboygan County, Wisconsin, where the Parcel Fabric — now also available in ArcGIS Online — was used to process a right-of-way dedication. Uploading a property deed into the COGO Reader tool automatically extracted point-of-beginning, distance, and direction data to construct the new parcel boundary, after which the Parcel Fabric automatically recalculated affected parcel areas and tax lines, and a parcel lineage tool traced how the property record had changed over time.
A second demonstration, using a proposed utility transmission line easement in Harris County, Texas, applied the same new file knowledge graph feature already seen in the transportation-centrality analysis — this time to analyze land-ownership networks among the 292 parcels within 500 feet of the proposed easement. The tool identified more complex corporate ownership structures behind some parcels, in one case tracing a single property owner’s connections through a general partnership to a total of 172 related parcels, allowing for consolidated stakeholder notifications rather than piecemeal outreach.
The two knowledge graph demonstrations were built from unrelated data — a road network in one case, a set of property deeds and corporate filings in the other — but they are the same underlying capability aimed at the same underlying problem: finding which nodes in a network of relationships actually matter, whether the network is physical infrastructure or legal ownership.
Enterprise at Every Scale
Elisa Coyle and Scott Rahauser, who together make up the entire GIS staff for the city of Allentown, Pennsylvania, described building dozens of applications for the city using ArcGIS Solutions templates, including a 311 citizen-reporting system adapted from the Citizen Reporter solution and a traffic sign inventory system built on street-view imagery and AI-assisted data extraction. In one example, the team repurposed a 5K fun-run route-tracking template to monitor the city’s litter-collection truck routes and pickup weights.
The afternoon session also opened with acknowledgment of the City of Austin’s GIS team, recipient of this year’s Enterprise GIS Award for more than 30 years of platform governance supporting roughly 14,000 city employees across some 30 departments — a scale of deployment that sits at the opposite end of the same platform from Allentown’s two-person operation.
Sam Libby returned to outline Esri’s Well-Architected Framework and Architecture Center, a two-year effort built with user feedback to document system patterns, reference architectures, and six architecture pillars for organizations building critical enterprise GIS systems. Libby also announced ArcGIS for ServiceNow, a new native, bidirectional integration allowing ServiceNow users to build ArcGIS-based map layers directly inside the ServiceNow experience and, in the other direction, bring live ServiceNow asset data into ArcGIS-powered applications and field workflows.
Hannajane Prichett and Megan Hendrick demonstrated new capabilities in ArcGIS Field Maps using a utility network model of San Diego-area electric transmission infrastructure. New features included branch versioning for managing concurrent edits to live utility network data, smart forms with multiple-choice fields and required-photo logic for on-site data collection, and downstream trace analysis to visualize electricity flow through a network. A separate demonstration showed Field Maps integrating directly with a handheld utility locator device, capturing pipe depth, orientation, and high-accuracy GNSS location data automatically as a field technician traced a buried natural gas line.
Representatives from NextEra Energy described how the company uses ArcGIS across the full lifecycle of energy infrastructure projects, from early-stage land suitability analysis and an in-house transmission-routing tool called Routify, which combines hundreds of data layers into a cost surface for evaluating route options, through construction monitoring with Field Maps and Experience Builder, to post-construction environmental stewardship. One example, an Experience Builder application called Wildlife Connect, supports environmental teams monitoring American crocodile nesting activity at the cooling canal system of the Turkey Point nuclear plant in South Florida. NextEra also described consolidating three separate ArcGIS Enterprise deployments into a single ArcGIS Enterprise on Kubernetes architecture to support a large internal GIS community it calls the Geospatial Squad.
At the far end of that same scale sits Rete Ferroviaria Italiana. Luigi Contestabile and Pier Paolo Olla of RFI, Italy’s national railway infrastructure manager, described a six-year GIS effort now supporting a network of roughly 10,000 miles of track and more than 2,000 stations, moving approximately 10,000 trains and over 2 million passengers daily. The company built more than 300 information layers describing customer catchment areas around each station, from housing and offices to universities and entertainment venues, allowing RFI to classify its 2,000 stations into six tiers based on the relevance of their surrounding urban context and to define station “vocations,” such as tourism or education, that guide which services and amenities are prioritized at each location. Contestabile described using this location intelligence for practical infrastructure decisions, including sizing bike parking capacity in the cycling-heavy city of Ferrara.
Pier Paolo Olla, who oversees day-to-day operations across RFI’s stations with a team of 1,200 colleagues and an annual budget of nearly half a billion dollars, described a subsequent indoor mapping effort that digitized 2,000 stations, more than 75 million square feet of indoor space, roughly 100,000 rooms, and 250,000 individually tracked assets in under two years, which the company describes as the largest indoor asset mapping project completed to date. That data now supports real-time maintenance dashboards and passenger accessibility applications showing which station areas are fully accessible, accessible via ramp or elevator, or not yet accessible, with plans to extend the system into personalized route guidance for passengers with mobility or visual impairments.
Between Allentown’s two staff and RFI’s 1,200, or between a 5K fun-run template repurposed for litter trucks and a continental rail network’s indoor digital twin, the platform underneath is nominally the same one. What differs is not the software but the organizational scale it has been asked to absorb — and the conference program placed both ends of that range on the same stage within a few hours of each other.
Digital Twins as Connective Layers
The San Francisco Airport and Vietnamese seaport twins introduced at the top of the afternoon program were not standalone showcases. Esri’s digital twin strategy lead, Christine Ma, framed digital twins generally not as standalone systems but as connective layers that unify operational, business, engineering, IoT, and geospatial data already held by an organization, built around three characteristics: integration of multiple data sources, time-enabled analysis of past, present, and future states, and an outcome-driven design tailored to specific decisions rather than one all-encompassing model.
Esri’s Madeline Schueren then demonstrated new 3D capabilities using a digital twin of Vancouver, Canada, including the new Google 3D basemap for photorealistic detail, Gaussian splats — a new highly realistic 3D layer type in ArcGIS built from drone-captured reality data, in this case supplied by partner Spexi — and new glow effects for illuminated features. A configured operational scenario showed a BIM model of BC Place Stadium integrated with emergency asset data for a prototype digital twin evaluated by Vancouver Fire Rescue Services for use during a major international sporting event, including 3D-enabled floor plans, fire asset locations, and viewshed analysis tools. A separate example showed crowd-simulation partner uCrowds using the same 3D foundation data to model pedestrian congestion patterns around a simulated festival near the stadium.
Disaster Response as the Throughline
The community survey had already introduced disaster response as a recurring theme; the afternoon’s award presentations supplied its clearest institutional examples.
Dangermond presented the President’s Award to the government of Jamaica, accepted by Prime Minister Andrew Holness. In his remarks, Holness described the response to Hurricane Melissa, which made landfall on Jamaica’s southern coast on the night of October 27-28, 2025, as the strongest storm ever to strike the island, with winds exceeding 185 miles per hour and storm surges reaching four meters, leaving more than three-quarters of the island without power. Holness credited three decades of institutional preparation, including a Land Information Council established in 1991 and a National Emergency Response GIS team formed in 2009, for enabling a single shared operating picture across government ministries, international partners, and organizations including the United Nations, Copernicus, NASA, NOAA, and ICEYE, coordinated in part through Esri’s Disaster Response Program — the same program Dangermond had cited earlier in the day as providing free software and content across hundreds of response efforts annually. Satellite-based damage assessment classified more than 960 buildings within days of the storm, informing decisions on shelter allocation, road clearance priority, and resource distribution. “A map is no longer just a picture on a screen,” Holness said. “It becomes the basis on which serious life-saving decisions are made.”
Esri’s Making a Difference Award this year went jointly to David Maidment and NOAA’s National Water Center. Maidment described the origins of his new book, Geowater — one of sixteen new titles from Esri Press launching this year’s Power of Where Collection, an extension of Dangermond’s 2024 book of the same name — connecting it to the 2014 death of Travis County, Texas, Deputy Sheriff Jessica Hollis in a flood near his home, an experience that shaped his subsequent push for real-time flood information. NOAA’s Fernando Salas described how the National Water Center used the Geowater framework as a blueprint to expand its forecasting from a few thousand points nationally to 3.6 million forecasts across 3.6 million river miles, implemented through the National Water Model and translated into neighborhood-scale inundation mapping used during events including Hurricane Helene, where Salas said flood inundation mapping data supported the evacuation of a hospital in eastern Tennessee without loss of life.
National Weather Service chief scientist Monica Youngman followed with an account of Hurricane Helene’s impact on her own community of Asheville, North Carolina, where the storm, beginning as a tropical depression on September 23, 2024, ultimately killed 170 people in western North Carolina and at least 250 across six states. Youngman described widespread and prolonged infrastructure loss in the region, including two to four weeks without power, cell service, or internet, and 54 days without water in some areas. She outlined two NWS-led projects addressing the gap between forecast data and public understanding of cascading risk: a Cascading Impacts project developed with MITRE and Esri that models dependencies between infrastructure systems, such as how many schools and students rely on a given water treatment plant, using a knowledge graph combined with weather and water forecasts; and Ask NWS, an in-development AI tool, discussed above, that helps forecasters translate authoritative weather data into plain-language, multilingual public safety messaging, reviewed and verified by a forecaster before release.
Read together rather than as two separate award presentations, Jamaica’s and NOAA’s stories are the same argument made at two different scales: Holness’s three decades of institutional GIS-building turned a Category 5 landfall into a coordinated, satellite-verified response within days; Salas and Youngman’s decade of forecast-model expansion turned a fixed number of river gauges into millions of location-specific flood forecasts precise enough to evacuate a specific hospital. Neither result was produced by a single tool introduced at this conference. Both were produced by institutions that had already spent years building the underlying system before the disaster that tested it arrived.
Education, Conservation, and the Long View
Dangermond closed the morning’s opening segment with an accounting of Esri as an organization, now 57 years since its 1969 founding by Dangermond and his wife, Laura. He cited more than 800,000 user trainings completed the previous year through Esri’s professional development programs, and highlighted 16 new titles from Esri Press this year, including Parcels: How Property Shapes Geography, co-authored by Regrid CEO Jerry Paffendorf and land-records systems expert Nancy von Meyer, alongside Maidment’s Geowater.
Dangermond also pointed to Esri’s business partner ecosystem, which he said now represents more than $35 billion in combined activity supporting Esri’s user base, and singled out three programs for particular acknowledgment: the disaster response program discussed above; the conservation program, which supports thousands of NGOs; and a decades-long education partnership with the National Geographic Society spanning K-12 and higher-education geospatial literacy initiatives.
National Geographic Society CEO Jill Tiefenthaler joined Dangermond to discuss the Society’s MapMaker platform, built in partnership with Esri and now reaching about 20 million students globally through a partnership with National Geographic Learning at Cengage. Tiefenthaler described MapMaker as designed for student-led exploration rather than teacher-directed instruction, allowing children to build their own map layers and pursue their own questions. The conversation also touched on the recently opened National Geographic Museum of Exploration in Washington, D.C., which includes the Laura and Jack Dangermond Esri GeoVerse, a 270-degree immersive exhibit space featuring GIS-based conservation stories, including efforts to reduce conflict between people and Andean bears in Peru’s cloud forest and to protect the wattled crane in Botswana’s Okavango Delta. Dangermond noted the partnership between Esri and the National Geographic Society, whose second president was Alexander Graham Bell, spans several decades.
The plenary’s closing keynote came from Kristine Tompkins, president and cofounder of Tompkins Conservation and former CEO of Patagonia, who described the organization’s origins in the early 1990s when she and her late husband, Doug Tompkins, co-founder of The North Face and Esprit, began acquiring land in Chile and Argentina. Over more than three decades, that work has helped create or expand 16 national parks — a 17th is set to be signed this coming October — protecting just under 16 million acres of land and more than 30 million acres of marine no-take zones. Tompkins described a shift in focus over time from land acquisition toward active species reintroduction, now working across roughly 35 species through the organization’s independent successor groups, Rewilding Chile and Rewilding Argentina, led respectively by Carolina Morgado and Sofía Heinonen. Reintroduced or recovering species include jaguars in Argentina’s Iberá National Park, Andean condors, giant river otters, and red and green macaws, work Tompkins said has not come without setbacks, including the loss last year of a young rewilded jaguar to poaching in the Gran Chaco, an incident that helped spur a national anti-poaching campaign in Argentina.
In a closing conversation with Dangermond, Tompkins described a push to expand the organization’s approach continentally, citing new engagement in Colombia, Brazil, and Ecuador, and described conservation and human development as inseparable goals rather than competing ones. “It’s people like you guys who have the possibility to stitch it back together again,” she told the audience. Dangermond closed the session by thanking attendees and inviting them to continue the conversation at the evening’s Map Gallery reception.