Someone in the field clicks a point on a phone, and a hundred people in an office watch the count fill in, live, during the three nights Los Angeles sends volunteers out to tally its unhoused population. That workflow, described inside a briefing the night before the Esri User Conference opened in San Diego, is the fastest growing kind of license the company sells. It is also the clearest picture of what geographic information system software does for an organization, which has almost nothing to do with the map you can see and almost everything to do with a decision that would otherwise get made blind.

Most enterprise buyers file this under mapping. That filing is the expensive mistake.

Two Esri leaders, Richard Cook from the services arm and his colleague running global business development, spent an evening walking a room of analysts through where the company sits after fifty-seven years. The technology detail mattered less than a pattern that kept surfacing: the value is not the picture, it is the answer to a question with the word where in it. Any industry with that question in its planning gains from the technology, and any team asking it without the data is guessing at a cost that arrives on a delay.

The bill for guessing shows up somewhere the decision-maker rarely looks

Break ground without an accurate map of what runs beneath it and the arithmetic gets brutal. Damage to buried utilities costs the U.S. economy at least thirty billion dollars a year, spread across more than 180,000 separate strikes, with a single incident averaging around fifty-six thousand dollars and a hit on a major transmission line running past a million (Common Ground Alliance, 2024). Marking and mapping failures, meaning the record was wrong or the asset was never documented, rank among the most persistent root causes, and their share is climbing, not falling.

The cost compounds the earlier it is ignored. Catch a utility conflict during design and the fix is a line moved on a drawing. Catch it during construction and the same change runs ten to a hundred times more expensive (Planning Tank, 2026). The map was always available. Someone chose to act before consulting it.

Retail runs the same equation above ground. More than seven thousand U.S. stores closed in 2024, a jump of nearly sixty percent over the prior year (Coresight, 2025). One analysis of retail failures found that sixty-seven percent of closed locations showed warning signs in their site data that nobody analyzed, or nobody read (GrowthFactor, 2025). A grocery store in a well-chosen trade area clears fifteen to twenty million dollars a year; the same format in a poorly read one struggles to reach half that. A wrong location traps five to ten years of lease, build-out, and staffing before an operator can walk away.

The cheaper substitute for the analysis is to follow someone who looks like they did it. Build wherever a Trader Joe's or a Starbucks already committed, the reasoning goes, because those companies spend heavily to pick sites and the storefront proves the location works. It is a real form of location intelligence, only a crude one. Starbucks screens for roughly 25,000 vehicles a day and heavy daytime commuter traffic; Trader Joe's optimizes for a dense, higher-income, higher-education grocery shopper. Inherit their address and you inherit a catchment tuned to their customer, not yours, priced up by the rent their arrival already pushed. Their decision is also two to five years old by the time you can read it off the street. Following it means signing a decade-long lease on a stale answer to a question you never asked in your own terms.

These are not exotic failures. They are ordinary decisions made a step too early, without the layer that would have flagged the risk.

The mine, the fleet, the pipeline all rest on the same question

Cook described the pattern across sectors that never announce themselves as GIS users. Energy and utility companies run the technology as a mission-critical system, not a mapping tool in the corner. A gas utility keeps one system for electric lines, another for gas, a wildfire management system, and a public safety power shutoff system, all married into maps through web services. A pipeline that fails usually fails at a segment where the record disagreed with the ground.

Fleet and transportation logistics turn on real-time location: where a vehicle is now, whether two units are converging on the same corner, whether traffic through an intersection has stalled. Asset management ties a physical valve or transformer to its record, its install date, its maintenance history, so a crew scanning a tag in the field pulls up the truth instead of a guess. Mining and materials discovery, retail siting, insurance exposure, defense, and fraud analysis all reduce to the same primitive. Location is a telling variable, and in fraud, waste, and abuse work it is often the most telling one.

Buyers underweight the scale under all of this. Esri's cloud platform alone streams four to five billion maps a day, and that figure excludes every organization running the technology on its own servers (Esri, 2026). Marriott monitors risk across roughly nine thousand properties on it. Heathrow is going live with a digital twin built on the same software. Most of the people relying on the output never say the word GIS.

The map is turning into something a model reads, not an eye

The briefing covered work Esri had not yet announced, so I am holding the specifics until they are public and will write about them after the announcement lands. The direction is safe to name because the company has said it in the open: a base map is becoming an input for machines, not only a picture for people. Layers of demographic, environmental, and movement data compressed into a form a model can query changes what the word map means inside an AI system. The map stops being the output and becomes the feature set.

That shift is the real reason a location layer is hard to fake. A large language model can produce a plausible sentence about a place it has never verified. It cannot produce the coordinate, the parcel boundary, the elevation, the flood line, or the buried asset unless something authoritative fed it that ground truth. The company that has spent decades curating who counts as authoritative is holding an asset the model layer above it cannot generate on its own.

The integration story is where the moat is being poured

A map is worth little if it cannot reach the systems where work happens. Esri has spent years wiring ArcGIS into the platforms enterprises already run, with Autodesk design data flowing both directions and ArcGIS maps living inside Microsoft Excel and Teams. The pattern is consistent: bring the map to where the decision already sits rather than pull the user into a separate GIS tool.

One analyst in the room pressed on the obvious gap. Esri visualizes physical assets and geography, but it does not hold the operational system of record, the enterprise resource planning data and process state that says what a facility is doing. Cook agreed and drew the line: the company will not build an ERP, it will partner with the vendors who own that data and connect to it. The bet is that a chief information officer wants to hear that two vendors already interoperate safely, not that one of them is trying to become the other.

That restraint is the strategy. It is also the exposure.