Our first warehouse robots were a disaster. A million-dollar disaster, to be precise.
I remember the day they arrived. Sleek, powerful, and smelling of fresh metal and future profits. We’d spent months hyping them up, convincing our investors and ourselves that this was the leap forward we needed. We were going to automate our entire fulfillment center, a complex dance of picking, packing, and shipping. These robots were the star performers, promising to 10x our efficiency and leave our competitors in the dust. On paper, it was a work of art. In reality, it was a slow-motion train wreck.
We didn’t see it at first. We were too blinded by the promise of cutting-edge technology. The idea was simple: a fleet of autonomous mobile robots (AMRs) would zip around the warehouse, bringing shelves of products to human pickers. No more walking miles of aisles. It was the classic “goods-to-person” model, the poster child for warehouse automation. We bought the best hardware, the most sophisticated software, and hired a team of brilliant engineers to stitch it all together.
Then came the first day of live testing. The first sign of trouble was subtle. A robot would just… stop. Right in the middle of a main thoroughfare, causing a traffic jam of other machines behind it. The engineers would rush over, plug in a laptop, and furiously type commands. “Just a minor software glitch,” they’d say. “We’re pushing a patch.”
But the glitches kept coming. One robot couldn’t seem to handle shiny objects. It would get stuck in a loop, trying to scan a roll of plastic wrap, its optical sensors completely baffled. We had to manually remove every reflective surface in its path. Another time, two robots tried to occupy the same physical space at the same time, resulting in a surprisingly loud and expensive crunch. It turned out their collision avoidance system had a blind spot when approaching each other at a perfect 45-degree angle. A one-in-a-million chance, we were told. Until it happened again a week later.
The real nightmare began when we tried to scale up to full capacity. The system collapsed. The central server coordinating the robots became overwhelmed, and the entire fleet froze. We had a warehouse full of orders, a team of pickers standing around with nothing to do, and a million dollars worth of metal statues. We had to revert to our old, manual process, and the team worked around the clock for three days straight just to clear the backlog. It was humiliating.
That’s when the miserable truth finally hit me. The robots weren’t the problem. We were.
We had fallen for the classic trap of thinking technology was a magic wand. We were so focused on the machines that we completely ignored the environment they were operating in. Our warehouse wasn’t a clean, predictable laboratory. It was a chaotic, ever-changing system of people, processes, and physical objects. A dropped pallet, a stray piece of cardboard, a puddle of water—these were things our human workers navigated without a second thought. For our robots, they were catastrophic, system-halting events.
Our biggest failure was not technical; it was a failure of imagination. We tried to impose a rigid, automated system onto a fluid, organic process. We didn't spend enough time understanding the real-world complexities of the warehouse floor. We didn't properly integrate the robots with our human workforce, creating a sense of frustration and mistrust. The workers saw the robots not as helpers, but as obstacles.
The decision to pull the plug was one of the hardest of my career. We had to write off the investment, a painful and public failure. But it taught me a lesson that has been worth far more than what we lost. It’s a lesson I carry with me in every investment I make today, especially in the world of robotics and physical AI.
Here’s what I learned:
Don't automate a broken process. If you have underlying issues in your workflow, adding robots will just make you fail faster and more expensively. We should have fixed our inventory management and layout issues before even thinking about automation.
The human-robot interface is everything. You can't just drop machines into a human environment and expect them to work. You need to design a system where humans and robots can collaborate effectively. This means intuitive interfaces, clear communication, and extensive training.
Start small, learn fast. Instead of trying to automate everything at once, we should have started with a small pilot project in a controlled area. This would have allowed us to learn about the real-world challenges and iterate on our solution without bringing the entire operation to a standstill.
That failure was a crucible. It forged my understanding of what it really takes to build successful companies at the intersection of the physical and digital worlds. Now, when I look at a pitch for a new humanoid robot or a drone AI company, I don't just look at the technology. I look for a deep, obsessive understanding of the problem they are trying to solve. I look for a team that has lived the chaos of the real world, not just the clean lines of a CAD drawing.
So yes, our first robots failed. They failed miserably. And it was the best thing that could have happened to me.
The Siren Song of Big Numbers
One of the biggest mistakes we made was getting seduced by the numbers on a spreadsheet. The ROI calculation was beautiful. It showed a 2-year payback period, a 50% reduction in labor costs, and a 3x increase in throughput. The numbers were so compelling that they created a reality distortion field. We started believing the spreadsheet more than the messy reality of our own warehouse.
I remember a board meeting where I presented the automation plan. I had a slide with a graph showing a steep upward curve labeled “Efficiency.” The investors loved it. We all loved it. We were high on the fumes of our own projections. What we didn’t have was a slide for “Complexity,” or “Unforeseen Problems,” or “Human Frustration.” We were measuring the wrong things.
We tracked robot uptime, pick rates, and battery life. We should have been tracking things like: how many times did a human have to intervene to fix a robot problem? What was the sentiment of the warehouse team towards their new robot colleagues? How much time was spent debugging software instead of shipping products? The data we were collecting was giving us a false sense of security, while the real problems were festering just below the surface.
The People Problem
I mentioned that the workers saw the robots as obstacles, but it was worse than that. They saw them as a threat. We did a terrible job of communicating the purpose of the automation. We said it was about “efficiency” and “cost reduction,” which everyone correctly translated to “we want to fire you.”
Fear and resentment grew. We started seeing small acts of sabotage. A strategically placed piece of tape over a sensor. A “lost” charging cable. Nothing major, but enough to add to the constant stream of small failures. We had created an adversarial relationship between our most valuable asset—our people—and our most expensive one—our robots.
Looking back, I should have made the team a partner in the automation process. We should have held town halls, asked for their input, and created a clear vision for how their jobs would evolve. We could have trained them to become robot technicians, fleet managers, or process improvement specialists. Instead, we treated them like cogs in a machine that we were trying to replace. It was a colossal failure of leadership.
The Autonomy Myth
The term “autonomous” is one of the most abused words in the tech industry. We were sold the dream of fully autonomous robots, a “lights-out” warehouse where the machines would run themselves. The reality is that we are still a long way from true autonomy, especially in complex, dynamic environments.
Our robots were great at following a pre-programmed path in a controlled environment. But the moment something unexpected happened, they were lost. They lacked the common sense and adaptability of a human. A human worker can see a spill on the floor and know to walk around it. A robot sees an unexpected obstacle and freezes, waiting for instructions. We spent millions on “autonomous” robots and ended up having to hire a team of people just to babysit them.
This experience has made me a huge skeptic of claims of full autonomy. I believe in a future where robots and humans work together, each playing to their strengths. Robots are great at repetitive, physically demanding tasks. Humans are great at problem-solving, creativity, and adapting to change. The magic happens when you combine the two, not when you try to replace one with the other.
The Aftermath and the Rebirth
After we pulled the plug on the robot project, the mood in the company was grim. We had failed, and we had failed publicly. But after the initial shock wore off, something interesting happened. The warehouse team, the same people who had been so resistant to the robots, came together with a renewed sense of purpose.
They started suggesting their own ideas for improving the warehouse. They created a new system for organizing inventory, a better process for picking and packing, and a more efficient layout for the fulfillment center. They were empowered, engaged, and motivated. Our efficiency didn't just recover; it surpassed the levels we had been chasing with the robots. We achieved our goals not with expensive technology, but with the ingenuity and hard work of our own people.
This was the most important lesson of all. Technology is a tool, not a solution. It can amplify the effectiveness of a well-run operation, but it can’t fix a broken one. The source of all value in a company is its people. My job as a leader and as an investor is to empower people, to give them the tools and the autonomy they need to do their best work. Sometimes that tool is a sophisticated robot. And sometimes, it’s just a whiteboard and a marker.
Today, when I visit the warehouses of companies I’ve invested in, I don’t just look at the robots. I look at the people. I watch how they interact with the technology. I listen to their conversations. I look for the spark of ownership and pride in their eyes. That’s where you find the real truth about a company’s future. Not in the polished metal of a machine, but in the spirit of the people who make it run.
Frequently Asked Questions
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.