BUILDING SHIT NOBODY ASKED FOR
Now with 300% more AI slop for investors.
I missed September's blog.
Fortunately, while I was away, the Earth Observation industry appears to have built another five platforms nobody asked for.
Another map. Another search bar. Another AI assistant that lets you type "show me Sydney" and then shows you Sydney.
Revolutionary stuff.
Chuck a few buzzwords into the investor deck, call it Geospatial Intelligence as a Service, raise $20 million and off you go.
Meanwhile, someone trying to buy satellite imagery is still emailing three different suppliers to figure out pricing, availability and when they can actually get the data.
But the new platform has dark mode, so that's something.
Are we actually solving anything?
Don't get me wrong. Making satellite imagery easier to find and purchase is a good thing.
Historically, this industry has made buying data unnecessarily difficult. Anything that simplifies that process is progress.
But we've somehow reached a point where putting another interface in front of satellite imagery is considered a technological breakthrough.
Search for imagery. View imagery. Buy imagery.
Great.
What happens next?
What happens when you need to task five different satellite operators? When the first satellite can't collect because of cloud cover? When the customer needs the data delivered directly into their own system, in a specific format, under their own security requirements?
What happens when you're processing thousands of requests rather than five?
That's where things get interesting.
It's also where a lot of the flashy product demos tend to end.
Now introducing AI. Because obviously.
It seems every second geospatial company has suddenly discovered artificial intelligence.
You can't just have a satellite imagery platform anymore. It needs to be an AI-powered geospatial intelligence ecosystem.
Preferably with an agent.
What does the agent do?
Well, you ask it a question about satellite imagery.
And then?
It tells you things about satellite imagery.
Fantastic.
I'm not against AI. There's enormous potential for machine learning, automated analysis, vessel prediction, change detection and a hundred other applications across Earth Observation.
But there's a pretty big difference between using AI to solve a difficult technical problem and bolting a chatbot onto an existing product because someone on the board asked about your AI strategy.
One solves a problem.
The other makes a good LinkedIn announcement.
Nobody wants another platform to log into
This is probably the bit we don't talk about enough.
Most customers don't actually want to spend their day buying satellite imagery.
A mining company wants to monitor its operations. An insurer wants to understand exposure. Defence wants to know what's happening in an area of interest. An environmental agency wants to understand what's changed.
The imagery is an input, not the end product.
Yet we keep building products around the assumption that customers want to log in, draw a polygon, compare satellite images and manually decide what to purchase.
Some do. Plenty don't.
They want the information to turn up where they already work.
If a system identifies something that needs investigating, why can't it automatically determine which satellites are available, arrange the collection, process the data and deliver it?
Why does someone need to open another website?
And why are we making the customer learn how the satellite industry works just to get an answer?
The boring shit nobody wants to build
Here's what's far less exciting than another AI-powered marketplace.
Connecting different satellite operators so their systems can actually talk to one another.
Normalising the data so customers aren't dealing with a dozen different formats.
Automating pricing, ordering, tasking and fulfilment.
Managing permissions, licensing, delivery and the inevitable exceptions when something doesn't go to plan.
Making all of that work across different providers, customers and infrastructure environments.
None of it makes for a particularly sexy fundraising announcement.
But if the industry wants to scale, somebody needs to do it.
We've got hundreds of satellites collecting extraordinary amounts of information about Earth. The technology in orbit is remarkable.
The processes for getting that information into the hands of people who need it are often surprisingly manual.
We've invested billions putting things into space. Perhaps we should spend a bit more effort on what happens when the data comes back down.
Making Earth machine-readable
This is where I think the bigger opportunity sits.
Not just making satellite imagery easier for humans to search and purchase, but making Earth Observation data accessible to machines and the systems that depend on it.
A system should be able to identify what data it needs, discover what's available, organise a collection, retrieve it and put it to work without somebody manually managing every step.
That's what making Earth machine-readable starts to look like.
And it's not just about imagery. It's about building the infrastructure to connect satellite data with other geospatial datasets, applications, analytics and automated workflows.
The industry has spent years making Earth easier to observe.
Now we need to make that information easier to use.
There's absolutely room for better marketplaces, better interfaces and better AI tools. All of these have their place.
But perhaps before we build another beautifully designed satellite imagery platform, we should ask a fairly simple question.
Who actually asked for this, and what problem are we solving?
Because if the answer involves a 40-slide investor deck, six mentions of generative AI and a demonstration that ends with somebody downloading a GeoTIFF...
We might have missed the point.