I started an AI automation company three weeks ago.
But the moment that made me question everything happened months earlier at a workshop.
A presenter was explaining how their AI assistants transformed procurement operations. Ninety people reduced to eighteen. She said it like a success story. And for the business, it was. Faster processing. Fewer errors. Massive cost savings.
I felt my shoulders drop. Physically drop. I couldn't hide it.
She noticed. Kept glancing at me for the rest of her presentation. I don't know what she saw on my face, but I know what I was thinking.
Seventy two people. Mortgages. Kids in school. Identities built around being good at jobs that suddenly don't need doing.
I still started the company. Here's why that's complicated.
The Pattern That Concerns Me Most
The pattern that concerns me most isn't mass unemployment. It's wage compression.
When AI does 80% of what a junior analyst does, that role doesn't vanish overnight. What happens is the remaining 20% doesn't justify the salary. One senior person supervises AI doing what five people used to do. The four displaced workers don't become homeless. They take worse jobs. They compete for positions that used to go to people with less education. And that pressure cascades downward.
We've seen this before with manufacturing automation, but it happened over decades and was geographically concentrated. This is faster and cuts across virtually every white collar function simultaneously.
The Shelf Life of "Learn to Use AI"
"Learn to use AI" is real advice right now. There's genuine leverage in knowing how to prompt effectively and chain tools together. But that advantage has a shelf life.
The tools get easier every release. Within a few years, using AI will be like using Google. Not a differentiator. Just the baseline.
What actually survives? Taste. Judgment. Relationships. The willingness to own outcomes when things break. Not because AI can't eventually do these things, but because they're the last to go.
What Happens to the Rest?
I don't have a satisfying answer for what happens to people who can't or won't continuously reinvent themselves. Some portion of the population doesn't have the cognitive flexibility, the resources, or frankly the interest. That's not a moral failing. It's just human variance.
Historically we've handled this through safety nets, make-work jobs, informal economies, and family support. I expect all of these to expand, unevenly and inadequately.
The political will for something like UBI doesn't exist in most places, and even where it does, the psychological damage of feeling useless isn't solved by a check.
So Why Build Automation Systems at All?
Because the alternative isn't that automation stops. It's that someone else builds it without the shoulder drop. Without the seventy two people crossing their mind.
The companies that will matter are the ones designing systems where humans stay in the loop not as a limitation but as the point.
Every customer-facing action my system generates requires human approval before it goes out. Not because AI can't write a good email. Because the relationship belongs to a person, not a process.
I'm not sure that's enough. But it's what I've got.
The Question I Can't Answer
What happens when "providing value through labor" becomes optional for a significant portion of people? We don't have cultural frameworks for that. Work isn't just economic. It's identity, structure, purpose, social belonging.
Some people will thrive finding meaning outside traditional employment. Many won't.
I don't know the answer. But I think the people building these tools should be asking the question out loud.
I'd rather be in the room where these systems get designed than watching from the outside while someone else decides what "efficiency" means.
This is part of our Building in Public series, where we share the real decisions, doubts, and questions behind running an AI automation company. If you're thinking about AI for your operations team and want to work with people who take these questions seriously, start a conversation.