Location data is embedded in almost every high-stakes decision a business makes. Where do you put a new facility? How do you route deliveries efficiently? Which markets are underserved? Where are your customers actually concentrated versus where you assume they are? These aren’t edge cases. They’re core questions that shape revenue, cost, and risk.
And for most of that decision-making history, the spatial dimension of those questions was either ignored or handed off to a specialist who answered it in isolation, without the full business context. The insight was locked in a department, and the decision happened somewhere else.
That arrangement made sense when location data required specialised tools, expensive software licences, and rare technical credentials to unlock. You staffed the expertise in one place and distributed the outputs. But the constraint that built that model has been removed, and when the constraint goes away, the model doesn’t just need to scale up. It needs to be rebuilt. AI is accelerating that rebuild faster than most organisations are ready for.
The constraint isn't access anymore. It's integrity.
What’s changed is not that location data has become more important. It has always been important. What’s changed is that AI systems, agents, and automated workflows are now consuming location data at a scale and speed that no human team ever did. Every time an agent makes a routing decision, a model scores a customer location, or an automated system triggers an action based on geography, it is drawing on whatever location data sits underneath it. The quality of that data is no longer just a logistics problem or a GIS team problem. It is an AI infrastructure problem.
Democratising access doesn’t automatically democratise quality. When location data was controlled by a small team, there were at least consistent processes for maintaining it. Now that it flows through more systems, more users, and more automated workflows, the duplication, the gaps, the outdated addresses, the mismatched boundaries, those problems scale too, and they scale quietly in ways that don’t surface until a decision goes wrong.
A delivery goes to the wrong address. A site gets selected based on foot traffic data that doesn’t reflect current patterns. An agent triggers the wrong action because the geographic boundary it was working from hasn’t been updated in two years. These aren’t hypothetical failures. They’re the kind of thing that happens when an organisation treats data access as the finish line rather than the starting point.
Location data is not just an address
When most people hear “location data,” they think addresses and coordinates. That’s the starting point, not the destination. What makes a location meaningful, to a human analyst or to an AI model, is everything that surrounds it: the demographic profile of the area, the businesses operating nearby, the land parcel boundaries and zoning classifications, the tax records tied to that property, the points of interest within a given radius, the historical patterns of activity associated with it.
All of those attributes are location data. And when an AI model is making decisions that have a geographic dimension, which increasingly means almost any decision involving customers, assets, or operations, it needs that full context, not just a coordinate pair. A model that knows where something is but not what surrounds it is working with an incomplete picture. At the speed and scale AI operates, that produces errors that compound before anyone catches them.
You can have a perfectly formatted, correctly geocoded address and still be missing the contextual layers that make that location useful for the decision at hand. Building that contextual richness into your location data, and keeping it current as boundaries shift, businesses open and close, demographics evolve, and tax records update, is the actual work. It doesn’t happen automatically when you adopt a new platform, and it doesn’t come for free with any AI tool.
The integrity question never goes away
At Precisely, we work on this every day, and I’ll be direct about what we’ve learned: location data degrades faster than most organisations expect. Geographic boundaries change, addresses change, and the contextual attributes tied to a location change too. What was true about a location two years ago may not be true today, and an AI model operating on stale data will reflect that staleness in its outputs.
The work of keeping location data trustworthy isn’t a one-time project. It’s continuous. You validate, you enrich, you eliminate duplicates, and you update for changes you didn’t anticipate. That work has to happen before the data reaches the decision, or the model, not after.
The organisations I’ve seen get the most out of location intelligence are not necessarily the ones with the most sophisticated tools. They’re the ones that built data integrity into the foundation first, with data that’s complete, accurate, contextually rich, and maintained over time. That foundation is what makes the AI layer trustworthy. Without it, you’re just automating on top of a shaky base.
The business case is harder to argue against than it looks
Location data errors don’t usually announce themselves. They show up as operational drag, as avoidable costs, as decisions that looked reasonable at the time, and as AI outputs that nobody can quite explain.
The upside case is real. Site selection done well, with clean spatial data and rich contextual attributes, changes where capital goes. Logistics routing built on accurate, current location data changes cost per delivery at scale. And AI systems built on a trustworthy location data foundation make better decisions, faster, with less human intervention required to catch and correct errors. These are the kind of outcomes that compound, and they become more pronounced as the AI layer takes on more of the decision-making load.
The question I'm still working through
As agents and models take on more decisions that have a geographic dimension, the question of who owns location data quality inside an organisation becomes more consequential. Is it centralised? Has it fragmented across the teams that use it most? Is anyone actually accountable for keeping the contextual layers current, not just the addresses? I think the ownership question is where a lot of the integrity problems start, and I don’t think there’s a settled answer yet.
Matt Waxman
Matt Waxman is chief product officer at Precisely, responsible for driving product strategy and innovation across the company’s product portfolio. With more than 20 years of leadership experience, Matt previously served as chief product officer at Arctera and held senior leadership roles at Veritas, Cohesity, Puppet, and Dell EMC, where he led product organisations delivering market-leading data management, cybersecurity, and cloud solutions.
