Spatial software vendors are sitting on a goldmine - and most don't know it yet

Alex Key | panoriq.ai

Nobody is better positioned for the agentic future than spatial software vendors.

Here’s what that future looks like – and it’s already working today: AI agents operating directly on point clouds, computing volumes from a terrain scan, or measuring rooms in a building scan and checking accessibility compliance. And perhaps most telling: users getting these tasks done by simply letting a standard tool like Claude write the code for them – tasks they would have needed dedicated software for until recently.

These were the demos in my talk at last month’s GEO Business Show, “Natural Language Meets Point Clouds”. But the real story isn’t natural language. It’s automation – and a shift that spatial software vendors should be paying close attention to right now.

You’ve probably noticed it in every industry by now: there’s an agent for this and an agent for that – agents that write code, review contracts, do the bookkeeping. And in all these industries, software vendors are panicking about the same thing: that their software is being replaced, either by users who now write their own tools, or by AI-native startups rebuilding their product in a weekend.

Here’s my take: the threat is real, but the panic is misdirected. And if you’re a spatial software vendor, you’re actually holding better cards than anyone else at the table. Let me explain.

First, what agents are not

Whenever I show agentic workflows on spatial data, someone asks: “So does this finally solve segmentation?”

It doesn’t. Everything gets jumbled up under “AI” these days, so it’s worth being precise. Agents are not new algorithms – they don’t suddenly solve classification or registration. And they’re not chatbots either. Nobody pays for chat; people pay for work getting done.

What agents are: automation for work that could never be automated before. You give an agent a goal and describe the outcome. It goes off, orchestrates your tools, writes code where needed, pulls in data, and comes back with the result – instead of a user clicking through the software step by step. In spatial data, I expect we’ll see a lot of hybrid versions, where the expert stays in the loop and collaborates with the agent.

That’s the shift: the interface becomes goal-oriented. Users care about outcomes, not about which menu item was clicked. And your software doesn’t get replaced by this – it gets operated. Agents sit on top of existing tools and orchestrate them.

Why vendors hold the best cards

Here’s the part that gets overlooked in all the panic. To build a great agent, you need three things:

  1. Real-life data. Agents only get good through iteration. Not on benchmarks, not on synthetic scenes – on actual messy production data.
  2. Domain expertise. In spatial data, this includes the algorithms. Working with spatial data is non-trivial, and those battle-tested pipelines are crystallized domain knowledge that a weekend of AI-assisted coding cannot replicate.
  3. Real users. An agent evolves through collaboration with the people who use it. You need users to work with, learn from, and improve the agent against their actual workflows.

A challenger startup has none of these on day one. Users building their own tools come closer – they have data and expertise – but they’re each solving their own problem, one workflow at a time. Nobody there is iterating a product against hundreds of users.

Established vendors have all three. That’s not a defensive position – it’s the best starting position in the race. The question is only whether you use it.

Why now?

Because the window is open, but it won’t stay open.

Challengers are already building. And your users aren’t waiting either: as I showed in my talk, general-purpose tools are already good enough to let them automate everyday tasks on their own. The spatial industry is a little behind others on this, but I don’t think it will stay that way for long.

Whoever starts the iteration loop first – agent, real users, real data, repeat – builds a lead that’s genuinely hard to copy. That compounding effect is the actual moat, not the feature list.

The good news: you don’t have to rebuild your product as an agent – most vendors won’t, and shouldn’t. Adding agent capabilities to your existing stack is a very realistic path. A first prototype that produces useful results on your own data takes weeks, not years – the volume computations and compliance checks I showed on stage are exactly this kind of prototype, and every vendor’s product is full of workflows like them. That early start is what matters, because it’s what kicks off that loop. Maturing it into something your customers rely on takes longer, which is exactly why the ones who start now will be ahead.

The companies that will lose aren’t the ones without an AI feature crammed into their product. They’re the ones that sit on the best data, the deepest expertise, and the most loyal users in the industry – and never turn those into an agent, while someone else builds the automation layer on top of their software. The cards don’t win the game on their own. You have to play them.

Alex builds AI agents for spatial data and helps spatial software makers bring agents into their products – more at https://panoriq.com. 

Thinking about what agents could mean for your software? Get in touch on https://www.linkedin.com/in/keyalex/ or at [email protected].

Our sponsors & partners

Thank you to all our sponsors and partners, who support the show throughout the year.

For more information about becoming a sponsor or partner, please email [email protected].

Lead Partners

Diversified Technology Portfolio

Organised by:

Diversified UK & Europe brings you the UK’s largest geospatial event for everyone involved in the capture, analysis and use of geospatial data.

Co-located with:

Privacy Preference Center