My Article in The Wall Street Journal

Today I want to share with you an article I wrote for The Wall Street Journal. The version they published — which you can read here — was edited for print space, so I’m sharing the full piece with you. 

The Robots That Handle Your Amazon Orders

A visit to a fulfillment center, where machines move merchandise and tell humans what to do with it.

When I had the chance to tour an Amazon fulfillment center in Denver with a few colleagues, I jumped at the opportunity. Like most people, I order regularly from Amazon but had never seen the machinery behind those quick deliveries. I was curious about the logistical marvel that could get a package from click to doorstep, sometimes in just a few hours.

What I expected was impressive efficiency and scale. What I found was both more remarkable and more unsettling than I had imagined.

The fulfillment center’s job is straightforward: receive goods from suppliers, sort and store them, then fulfill customer orders by packaging items and loading them onto trucks. This particular center wasn’t responsible for final delivery – it sends already-labeled packages to smaller distribution centers closer to customers, which hands them off to carriers like UPS, USPS, or Amazon’s own delivery fleet.

The first thing that hits you is the noise. The conveyor belts running through the facility are so loud that we needed headsets to protect our hearing. The building itself is enormous — three stories high, with about 2.4 million square feet of working space, roughly the size of sixty football fields. It’s divided into three distinct areas. The perimeter is a lit walkway that allows goods to flow in and out of the building. About one-third of the center looks like a normal warehouse where humans work.

But the largest portion of the facility is completely dark. This is where inventory is stored on thousands of yellow shelves, each five feet square at the base and eight feet tall. The robots swarming this land of darkness look like electric orange self-driving lawn mowers in a coordinated dance, giving each other the right-of-way. They slide under shelves, lift them a few inches, and ferry them around the dark area. A fence separates the dark zone from the rest of the facility, protecting the humans from the robots and the inventory from humans. Only specially trained employees can enter when equipment malfunctions or an item falls off a shelf.

Chaos Theory in Practice

Unlike at Home Depot (or most retailers), where knives are with knives and bolts are with bolts, almost all the inventory here is stored on shelves scattered throughout the dark zones in a completely random order. Amazon’s computers keep track of millions of items across thousands of shelves – they know not just which shelf contains each item, but exactly where it’s placed on that shelf.

All human interaction with the shelves happens at the “DMZ” – the boundary between the light and dark zones, where the loading and picking stations are located. At loading stations, merchandise is taken from yellow bins and placed onto shelves for storage. At picking stations, merchandise is retrieved from shelves and put into yellow bins.

Both stations look similar, and there are hundreds of them. Each is staffed by one employee who follows pictorial instructions on several overhead screens. For loading, they might see a visual instructing them to take an item from bin #2 and place it on the right side of shelf level five. If they are picking, similar visual instructions might tell them to take the item pictured from the middle of shelf level four and put it into bin #5. A light shines on the correct one of the five numbered bins in front of them.

Multiple cameras monitor each employee’s movements, recording where each item is placed. Computers track the time it takes to load or unload each item, and a single screen displays the employee’s statistics. Amazon has gamified the process so that pickers and loaders can compete against each other using their times as inputs. Every hour, the computer interrupts the workflow to provide short, guided stretching exercises.

As I talked about this distribution center with my colleagues, I remarked that the dark side of the fence where the robots roam looked dystopian, like something out of Black Mirror. But my colleague Robert offered a more chilling assessment: “It’s the humans that look dystopian.”

He was right. Workers exercise zero judgment; they simply follow detailed computer instructions. As I observed them, my heart overflowed with empathy. They are shells of muscle, mindlessly moving items from one place to another. This job requires someone to turn off their brain for 10 hours a day and enter a comatose state for the duration of their shift (they work four 10-hour shifts a week).

This job starts at $22 per hour – a relatively high wage, considering Colorado’s minimum wage is about $15. A job at 7-Eleven starts at minimum wage, and entry-level positions at Walmart pay about $18. But this job requires wearing protective headphones for 10 hours a day in a windowless building in near-total isolation. It’s just you and the yellow bins and shelves, without interacting with another human for almost the entire day.

As we discussed automation with our guide, we learned this facility was a “Version 9” – a relatively old 2015 vintage (though I imagine the software running the automations is days old). The newest fulfillment centers are up to Version 13, with even more robots. It’s easy to predict how this story will end – these distribution centers will literally “go dark,” as robots don’t need light. It’s just a question of whether it will happen in five years or ten.

Then I realized what I had initially missed: this facility is run by sophisticated algorithms housed in an AWS facility a thousand miles away. Every single decision, from the moment an order is placed, is made by these algorithms—not AI, but millions of lines of pre-programmed code. A software system decides which facility to send the order to. Once the order arrives, the lawnmower-like robots start moving shelves. The system then provides pictorial instructions to the humans on what to pick, where to place it, and which truck to load.

Thousands of micro-decisions were made by algorithms, and none by humans. The people here (outside of those who designed the system) were just cogs – biological components executing software instructions with remarkable precision, but without thought, creativity, or autonomy.

The future hasn’t arrived yet, but its foreshadowing is alive and well, humming quietly in windowless buildings across the country, where the boundary between what humans can do and what machines can do grows thinner every day.

For well over a century labor and capital have been at war.  Automation and sophisticated algorithms have been steadily winning for capital. But AI—which learns from data and improves itself rather than simply following pre-programmed rules— is an accelerant that will bring  decisive victory to capital. Robots will perform almost all tasks; humans will be there only to troubleshoot edge cases. Today, this facility employs 3,100 people. Fully automated, combining robotics with AI-driven coordination, it may need only 100.

I honestly don’t know how I feel about it. On one hand, 3,000 people will no longer have to turn themselves into drones for 40 hours a week. On the other hand, what will they do? What will happen to millions of people like them across the country?

This question haunted me as I left. A common argument, which I have often made myself, is that AI and automation will create new, high-paying jobs, just as previous technological revolutions did. This perspective has historical precedent: a century ago, one-third of the U.S. population worked on farms before mechanization shifted that labor into factories. Similarly, many high-paying occupations today (cloud engineer, social media manager, SEO expert) didn’t even exist thirty years ago.

Unlike those past shifts, however, the AI revolution may be so pervasive – broadly impacting both blue-collar jobs, from robots manning warehouses to self-driving technology replacing drivers, and high-paying white-collar jobs like radiologists, accountants, and customer service reps – that it eliminates jobs without creating enough new ones. Past revolutions also took decades to unfold. The question isn’t if new jobs will emerge – they will. The question is whether they will emerge fast enough, and in sufficient numbers, to offset those displaced by AI.

We’re likely entering an era of structurally higher unemployment. Walmart has already announced that it will stop increasing its workforce, despite growing and opening new stores. Companies today are walking a thin line – they don’t want to appear to disturb the perception of the status quo, but automation and AI will lead to a decreasing need for human workers. Hiring freezes will come first, followed by layoffs.

The consequences for society are truly difficult to imagine. The natural reaction will probably be to embrace populism. Corporations will become social villains. Will we have political unrest? Mass unemployment combined with wealth concentration in the hands of those who own the algorithms? Nobody knows. But the world that lies ahead will be fundamentally different from what we’re used to.

Watching these workers reduced to biological automatons made me think about my own children. What skills will they need to thrive—not just survive—in this new world?  They’ll need creativity – not just the artistic kind, but the ability to connect disparate ideas in new ways. Higher emotional intelligence and human skills will become paramount as more work moves into algorithmic domains. They’ll need adaptability – the ability to constantly acquire new skills, as they’ll have multiple careers over their lifetime in a world changing at a blistering pace. They’ll need to maintain agency over their lives rather than develop a victim mentality when faced with technological disruption. They’ll need to stay curious, embracing AI as a tool to leverage their human capacity, and remain open to change. They will, in short, need to cultivate the very things that make them irreducibly human.

Machines will stuff boxes and follow instructions. My hope is that my children will write their own script rather than read from one.

Key takeaways

• Touring the Denver fulfillment center revealed the choreography of Amazon orders robots, moving shelves in perfect harmony through a vast, dark space while humans stood at the edges, following instructions from glowing screens.
• What looks like technological progress is also a quiet moral trade: humans reduced to executors of algorithms, paid decently but required to suppress judgment and creativity for ten-hour stretches in isolation.
• The real intelligence in these warehouses isn’t artificial yet—, it’s algorithmic. Every item picked, every bin loaded, every truck dispatched is dictated by software, not by human decision-making.
• As AI evolves beyond pre-programmed rules, these already dimly lit centers will literally go dark, needing only a fraction of today’s workforce. Efficiency will rise, but at what human cost?
• The lesson extends far beyond logistics. As machines take over the mechanical, we’ll need to double down on the human: curiosity, creativity, emotional intelligence, and the courage to keep writing our own scripts instead of reading from someone else’s.