The Company That Didn't Need Us Anymore
My parents started the company in their garage. By the time I became CEO, we employed 140 people. We manufactured specialized components for commercial buildings. It wasn't glamorous. It was ours. Then AI changed almost everything. Design became automated. Procurement became predictive. Robotic manufacturing became affordable even for companies our size. AI agents negotiated with suppliers and customers. Competitors that once needed 100 employees could operate with 30. Within five years, I faced a fact I didn't want to admit: Our company could probably produce more with fewer than half our people.
The financial answer was easy
Automate aggressively. Reduce headcount. Increase margins. Remain competitive. The models were compelling. If I waited too long, the entire company might fail. Then everyone would lose their jobs. But these weren't numbers. I knew these people. Some had worked for my father. Their children went to school with mine. I wanted HOS to tell me the responsible thing to do. It couldn't. There were responsibilities pulling in different directions. Employees. Customers. My family. Shareholders. The community. Future employees. And my own responsibility to keep the business viable. Respecting people's agency didn't mean guaranteeing everyone a job forever. But it also didn't mean pretending that efficiency was the only value involved.
I delayed
That was my first major mistake. I told myself I was protecting people. Really, I was avoiding the decision. Competitors moved faster. We lost two major customers. Suddenly we had less flexibility and less money available to help employees transition. My delay had consequences. That was painful to admit. Sometimes refusing to choose is still a choice.
We changed the question
Instead of asking only: “How many jobs can automation eliminate?” we began asking: “What company are we trying to become, and what human capabilities will it require?” Some jobs disappeared. There was no way around that. But others changed. Experienced production workers became automation supervisors. Customer-service employees became problem-solvers for complex accounts. Some people retrained successfully. Some didn't want to. One of our best employees told me: “I've done this for 28 years. I don't want to learn how to supervise robots.” I wanted to convince him. Then I realized I was doing exactly what I said I didn't want technology to do to people. He had agency too. We created a transition package.
He left. I hated losing him. He was happy. Those two things could both be true.
AI eventually became part of management
Years later, AI systems participated in almost every significant company decision. They modeled demand. Suggested capital investments. Identified performance problems. Predicted which employees might leave. Recommended compensation. They were often right. That created another temptation. If the system could predict that an employee had an 82 percent probability of leaving, should we stop investing in that person? If it predicted someone was unlikely to succeed in management, should that person ever get the opportunity? If it knew an employee was struggling financially or emotionally from personal data, should the company use that knowledge? The technology could know more. That did not mean the company was entitled to know everything it could know. We established boundaries.
Some reduced efficiency. I came to believe that was acceptable. Human beings weren't merely inputs into the company's optimization model.
Then the company nearly failed anyway
A technology shift made our primary product less important. For all our planning, we had missed it. Revenue collapsed. We had to make another painful reduction. I felt as though I had failed my parents, our employees and myself. HOS helped me examine the failure without rewriting history. Some decisions had been good. Some hadn't. Some circumstances could reasonably have been anticipated. Others couldn't. Resilience did not mean pretending every failure contained a hidden victory. Sometimes failure was simply loss. The question was what we did next. We survived. Smaller. Different. Eventually stronger.
Twenty years later
The company employed only 85 people. It produced nearly ten times what it had when I became CEO. That statistic bothered me for years. Then I learned to see the fuller picture. Many jobs had disappeared. Other kinds of work had emerged. Some former employees started businesses of their own using technologies that once would have required millions of dollars. One became a competitor. I wasn't thrilled about that. I was proud of her anyway. Technology had concentrated extraordinary capability. It had also distributed capabilities once available only to large companies. Both things were happening simultaneously.
Looking back
I originally thought HOS would help me make better business decisions. It did. But it also kept forcing me to remember something business systems can easily forget. Every employee had a life beyond our company. Every customer had interests beyond ours. Every decision affected people who had their own purposes. That didn't mean avoiding hard decisions. I eliminated jobs. I closed a facility. I made decisions people hated. Some were right. Some were wrong. Some I still don't know how to judge. HOS didn't turn business into charity. It helped me make decisions with a fuller understanding of whose agency was affected and what responsibilities I was willing to accept.
AI made our company extraordinarily capable. But capability alone never answered the question: What should we do with that capability? Humans still had to answer that. I still had to answer it. And accepting responsibility for the answer became one of the most important parts of my life.
Here the individual using HOS has power over other people's lives. That changes the nature of the responsibility without changing the underlying principle. AI can make the company vastly more productive, predict employee behavior, recommend workforce reductions and identify increasingly efficient decisions. HOS does not tell the owner to reject those capabilities, nor does respect for agency mean preserving every job or avoiding difficult business decisions. Instead, it helps her see more completely who is affected, what responsibilities accompany her authority and where technological capability may exceed legitimate use. Her delay in confronting automation demonstrates that good intentions do not remove consequences; avoiding a difficult decision can itself diminish future choices. Later, the company deliberately declines to use some information it could obtain about employees. This illustrates an important distinction for a future HOS: the ability to know, predict or influence something does not automatically create the right to do so. Another AI might define success primarily through the company's objective and optimize accordingly. HOS must understand the objective while also helping the individual recognize the equal humanity and agency of those affected by pursuing it. Greater capability therefore brings not only greater opportunity, but greater responsibility for how that capability is used.