I Have 23 Employees, None Are Human: My AI Workforce

Mama, how many people work for you at The Change Republic? Twenty-three, I said.

My son looked at me. He knows my office. He has never seen twenty-three people walking around. There is me, my laptop, and a lot of coffee. So he asked the obvious next question. Then where do they sit? They sit inside my computer. And they work 24/7. Let me explain what that actually means. This is not science fiction, and you do not need to be technical to copy any of it.

Everything runs in Claude Code, a no-code environment where multiple AI specialists operate across eight teams. The build is straightforward because the thinking part requires a human. I wrote instructions in plain English, describing how I work, what good looks like, and what I never want done. Each of the twenty-three is a small AI worker doing one job well, repeatedly and consistently.

Klaus serves as my chief of staff, the one I speak with all day. Rather than managing twenty-three workers directly, I communicate with Klaus, who determines which specialist should handle each task and ensures completion. I keep conversations simple. Klaus, sort my expenses. Or, Klaus, what should I be paying attention to this week?

The system works because I split repetitive jobs across specialists happy to do one thing forever and do it well. Klaus connects them into coherent work, similar to how any company operates with specialists going deep on individual tasks and one good manager connecting them.

My team includes a writer who captures my voice, a video editor who reshapes content for different platforms, a researcher who provides weekly AI and leadership summaries, a website traffic analyst, and a quality control reviewer. When I face difficult decisions, another specialist presents arguments from multiple angles before I choose.

The expenses worker exemplifies this approach. I screenshot pending expenses, hand the list to Klaus, who passes it to the specialist. The worker logs into apps, downloads tickets and receipts, and organises everything in one folder. Tasks that used to consume entire evenings now finish in the time it takes me to make coffee.

I provided login credentials intentionally, because the worker needs them to perform the work. The one boundary I maintain: I handle the final bank upload myself. The team manages the work. I keep money in my hands.

A year ago, I managed everything alone. Now the boring parts happen automatically. Repetitive tasks run overnight. I arrive to find work already progressing while I slept.

The secret to starting is simple. Pick the task you dread most and teach an AI tool how you do it once, in your own words. Begin with one. Build additional specialists as needs emerge. Eventually, you realise you need a Klaus to coordinate them.

This philosophy extends to my professional practice. I coach leaders and teams on AI leadership, exploring the same questions I asked myself. What gets handed to AI, what stays with humans, and where to draw the line.

Many focus on the jobs AI eliminates. I focus on the work I handed over intentionally. None of my specialists coach clients, speak to audiences, or make in-the-moment decisions about what people need. That remains mine. They handle chasing, drafting, cutting, reminding, and checking.

My son asked where my twenty-three employees sit. They sit in the background doing what I do not need to do myself, so my hands stay free for work only a human can do.

If you could hand one job to a worker like that tomorrow, which would you give away first?

I first shared this as a LinkedIn newsletter, where you can see the visuals of how the team is set up. You can read the original version with the images here: https://www.linkedin.com/pulse/i-have-23-employees-none-them-human-t%C3%BCnde-lukacs-stnue/

Want to think through what AI means for your own leadership, and where to draw your own line between machine and human? My Private AI Advisory is built for exactly that.

https://www.thechangerepublic.com/private-ai