How geography enables better AI outcomes.

Artificial Intelligence (AI) has moved from a specialist research field into a mainstream business technology in a remarkably short space of time. Across the island of Ireland, organisations are wrestling with what that means for them. They are creating roles, drafting policies, experimenting with use cases and running pilots, and the pace of adoption has been striking. Everyone accepts that AI can reason but, on its own, can AI reason accurately, if it doesn’t know the spatial context in which it is making decisions?

Most AI programmes rest on a simple assumption: give a model enough text and enough tables, and it will provide the correct answer. Unfortunately, that assumption is flawed. Models can generate responses that are fluent, confident and plausible, but still wrong or entirely fabricated. This phenomenon, now widely known as hallucination, occurs because a model predicts what a convincing answer should look like rather than retrieving a verified one.

The problem is not that AI lacks intelligence. The problem is that it often lacks context. A language model can summarise a planning report, review a development policy document or reconcile two site survey spreadsheets very capably. However, it cannot tell you on its own that the site sits within a Special Area of Conservation, that it lies three hundred metres from a wastewater treatment plant already at capacity, or that the only road serving it floods every winter.

Those facts are held in a map, not in text, and that gap is why geography needs to move from a specialist discipline to something every organisational AI strategy depends upon. Reasoning needs reliable facts to stand upon, and for many organisations those facts are fundamentally about place. Geography provides the organising framework, the unique index, that links information together and gives meaning to otherwise disconnected datasets.

The point is less about mapping and more about integration. Location is the one common field capable of linking data that nothing else can link. A utility asset register, a census small area and a flood extent have little in common until they are placed on a map and viewed in context.  

AI models are extremely good at recognising patterns and remarkably poor at recognising what they are missing. They do not know what they do not know. An AI agent asked to prioritise maintenance across a network will generate an answer whether or not it understands local ground conditions, asset criticality, ownership boundaries or environmental constraints. Without context, the recommendation may be plausible but unreliable.

Spatial data fundamentally changes this dynamic. It narrows the range of possible answers and introduces evidence-based constraints that help separate actionable recommendations from those that simply read well.  This becomes even more important as organisations move towards autonomous and semi-autonomous systems. If AI is going to support decisions about infrastructure investment, housing delivery, environmental management or emergency response, it needs access to authoritative information about the real world.

At Esri Ireland, this argument formed a key part of our Esri Ireland Conference 2026 discussions. We described Esri’s geographic information system (GIS) technology, ArcGIS, as the platform that gives AI a trusted picture of what is happening on the ground rather than one that merely sounds plausible.  As organisations increasingly seek to automate analysis and decision support, geographic data provides the factual foundation that allows AI systems to operate with greater confidence and accuracy.

Geography provides context, while also supporting transparency and auditability. Decisions can be traced back to specific datasets, locations and conditions, creating a clearer record of what the system knew at the time a recommendation was made.

The island of Ireland has considerable AI ambitions and is also particularly well placed to benefit from AI.  It is small enough that national datasets can genuinely provide broad coverage, yet diverse enough that local conditions often determine the right answer.

Geography, though, remains the key.  Many of our island’s most pressing challenges are fundamentally geographic. Housing delivery, water and wastewater capacity, grid connection availability, transport investment, climate adaptation and flood resilience are ultimately questions about place.  The good news is that much of the underlying data to address these challenges already exists. Central government departments, local authorities, utilities, environmental bodies and transport agencies have invested significantly in high-quality spatial information.

The challenge organisations most often encounter is not a lack of data but a lack of preparation. Many have not yet structured, documented and published their data in ways that make it readily usable by AI systems.  This is gradually changing, in part because of the regulatory environment.  Authoritative spatial data aligns naturally with the requirements emerging from frameworks such as the EU AI Act. 

AI is rapidly expanding in use right across the island of Ireland.  In time, it will be relied upon to help inform many of the big decisions that will impact our lives, the built infrastructure we depend on and the natural environment we cherish.  Our critical purpose now is to make sure that AI is making these big decisions intelligently based on the right information, which undoubtedly includes the spatial context provided by geographic information.  As much as we increasingly need AI to help us solve national, local and organisational challenges, AI needs geography.