The Step Most Warehouses Skip Before Deploying Robots

In the spring of 2025, a logistics director at a large Nordic retailer did something that would have seemed odd to anyone watching. He pulled out his phone, opened a stopwatch, and followed one of his warehouse workers around for an entire shift.

The worker's name doesn't matter. What matters is what the logistics director counted: fourteen times that day, the worker left what he was doing to go look for something. A pallet that had arrived but hadn't been racked. A shipment that the system said was in aisle 12 but wasn't. A product that a customer needed and nobody could locate without walking the floor.

Fourteen searches. Average time per search: about twenty minutes. That's nearly five hours — in a single shift, in a single building — of a trained worker doing nothing but walking around looking for things that a computer somewhere claimed to know the location of.

The logistics director did the math. Across his three facilities, he estimated his company was burning something like fifteen thousand hours a year on searching. Not picking. Not packing. Not shipping. Searching.

"The crazy thing," he said later, "is that we had a warehouse management system. We had an ERP. We had dashboards. We had everything you're supposed to have. And none of it could tell me where a pallet actually was."

The Fiction of Perfect Information

This is a story about a gap — a specific, expensive, and strangely persistent gap — between what warehouse software believes is true and what is actually true on the floor.

Every modern warehouse runs on systems designed to track inventory. A WMS records transactions: something was scanned here, moved there, shipped out at this time. An ERP connects those movements to orders and forecasts. Together, they create a picture of the operation that looks, on a screen, clean and complete.

The problem is that the picture is a fiction. A useful fiction, mostly. But a fiction.

The WMS knows that a pallet was scanned at the dock at 2:14 PM. It does not know that the pallet then sat on the dock for three hours because the racks were full and nobody told the system. It does not know that someone moved it by hand to a temporary spot in aisle 9 that doesn't exist in the software. It does not know that aisle 9 is now blocked, which means the forklift driver who needs to access aisle 10 will take a four-minute detour, fourteen times today.

The systems track transactions. Nobody tracks the space.

What If Your Warehouse Could See Itself?

There is a term for what fills this gap, and it sounds, at first, like the kind of phrase a consulting firm would invent to sell a PowerPoint: spatial intelligence.

But the concept behind it is disarmingly simple. What if your warehouse could see itself?

Not through barcode scans or RFID tags or manual cycle counts. Through vision. A camera — mounted on a forklift, carried by hand, or strapped to a robot you already own — moves through the facility. Software turns what it sees into a three-dimensional model. Not a blueprint. Not a schematic. A living, searchable, queryable representation of what is actually on the floor right now.

You log in. You type: "Where are the blue pallets from the Tuesday shipment?" The system highlights them on a 3D map. Three seconds. Not twenty minutes of walking. Three seconds.

The logistics director, when he first saw this, said something that might sound like an exaggeration but probably isn't: "It was like putting on glasses for the first time."

The Immediate Economics

The immediate economics are almost embarrassingly straightforward.

The search problem — fourteen walks a day, twenty minutes each — largely disappears. The space problem, where a typical warehouse runs at 68% utilization because nobody can actually see where capacity is hiding, becomes visible overnight. A heatmap shows you that Zone B is crammed and Zone D is half-empty, and suddenly you've found 15 to 25% more usable space without leasing a single additional square meter. The audit problem — thirty-two hours a month of workers counting inventory instead of moving it — shrinks by 80%, because the system already knows what changed since yesterday.

Add those up for a mid-sized warehouse and you land somewhere between €200,000 and €500,000 a year. That's the value of the map alone. And this is where the story gets interesting.

The Robot Deployment Problem

Because there is a second thing happening in warehouses right now, and it intersects with the first in a way that almost nobody has noticed.

The robotics industry has spent the last decade building increasingly capable autonomous machines. AMRs that can move pallets. Forklifts that drive themselves. Picking arms with machine vision. The hardware works. The software works. What doesn't work — what keeps failing in pilot after pilot — is the deployment.

A robotics executive, who asked not to be named because he was being unusually honest, put it this way: "We can build a robot that navigates a warehouse perfectly. The problem is, nobody can tell the robot what the warehouse looks like."

Think about that for a moment. We've built machines that can navigate autonomously, but we're deploying them into environments they can't see. Every robot vendor, when they arrive for a pilot, has to build their own map of the facility from scratch. That's why pilots take months to set up, work in only one zone, and break the moment someone moves the racking.

This is the insight that the spatial intelligence story hinges on:

The same 3D model that tells a human where the blue pallets are can tell a robot where to go.

A robot operating on a spatial intelligence layer doesn't need its own mapping system. It inherits one. It knows where things are, which paths are clear, where congestion is building — on day one, in every zone, regardless of which company manufactured it.

The map comes first. The robots come second. Nearly every warehouse that has struggled with automation did it the other way around.

Start Small, Scale Fast

What makes this particularly interesting from a boardroom perspective is the staging.

You don't have to believe in autonomous warehouses to benefit. You don't have to buy a robot. You don't have to commit to anything beyond a 90-day test.

A company exploring spatial intelligence typically starts with a single site. Someone walks the floor with a camera. Within 48 hours, the facility exists as a searchable 3D model. Over the first month, the system builds a baseline — where space is wasted, how long searches take, where bottlenecks form. By the end of the third month, there's enough data to answer a simple question: is this valuable or not?

The cost sits within an operations budget. There's no procurement cycle, no IT integration, no infrastructure to install. And because the map delivers value immediately — through search elimination, space recovery, and audit reduction — the pilot tends to pay for itself while it's still running.

The companies that move fastest tend to be the ones that frame it not as a technology purchase but as a question: what can't we currently see?

If the answer is interesting — and so far, it has been — then the map is already built. The robots, whenever they come, will have somewhere to go.

Seeing for the First Time

The logistics director from the Nordic retailer ran his pilot. Within six weeks, his team had stopped searching. Not reduced searching. Stopped. The utilization data showed 22% more usable space than anyone had estimated. The cycle count hours dropped by four-fifths.

When asked what surprised him most, he didn't mention the savings. He mentioned something else.

"We thought we knew our warehouse," he said. "We'd worked there for years. But we'd never actually seen it before."