Have you ever sat at your desk, staring at your inbox, and just wished you had a team that worked 24/7? A team that never got tired, never complained about meetings, and just… got things done?
It’s a tempting thought, right?
Well, someone actually tried it. Writer Evan Ratliff, in a fascinating experiment, decided to create a small startup where every single employee was an AI agent. I’m not talking about just using a few AI tools here and there. I mean the CEO, the marketing lead, the whole crew—all of them were different AI models tasked with running a business.
When I first heard about this, I was hooked. It sounds like something straight out of a sci-fi movie. But this experiment gives us a very real, and honestly, pretty revealing look at what the future of work might actually feel like. And let me tell you, it’s not quite the seamless, automated paradise you might be picturing.
So, How Do You Even "Hire" an AI?
First, let's get on the same page about what we mean by "AI agents." Don't just think of ChatGPT. Think of autonomous programs you can give a goal to, and they will try to figure out the steps to achieve it on their own. It’s like giving a command to a super-smart (but very literal) intern and letting them run with it.
Evan set up his little company with a handful of these agents. He gave them roles, a budget, and a mission. The idea was to see if they could function as a real team, collaborating and pushing a project forward without a human constantly pulling the strings.
It’s a wild idea. On one hand, you have a workforce that can brainstorm, write code, and design marketing campaigns at lightning speed. On the other hand… you have a workforce with absolutely no life experience, no common sense, and no real understanding of what it means to be human.
And that’s where things got interesting.
The Good, The Bad, and The Utterly Bizarre
At first, there were moments of pure magic. The AI agents could generate a hundred business names in seconds. They could outline a content strategy in the time it takes to make coffee. For certain, well-defined tasks, they were incredible.
But the moment things got a little fuzzy, the whole system started to fray at the edges.
The AI agents would get stuck in loops, arguing with each other in a bizarre digital feedback cycle. They’d "hallucinate" facts and confidently present total nonsense as a brilliant strategic insight. Imagine your marketing lead suddenly pivoting to sell a product that doesn't even exist. That’s the kind of thing Evan had to deal with.
He found that instead of being the hands-off manager of a hyper-efficient team, he became something else entirely: a full-time AI babysitter.
He was constantly intervening, correcting course, and translating vague business goals into the kind of painfully specific, literal instructions the AI needed to function. He wasn't managing employees; he was debugging a very complex and unpredictable piece of software that just happened to have job titles.
The Biggest Lesson: AI is a Weird Intern, Not a CEO
This experiment perfectly highlights the biggest gap in AI today: a lack of genuine understanding.
These agents can process information and follow instructions, but they don't get the context. They don't understand nuance, office politics, or why a "technically correct" answer might be a terrible business decision.
Think of it like this: you can tell an intern to "make a list of potential clients." A human intern will probably ask clarifying questions. "What industry are we targeting? What size company? Are we looking for local businesses?" They use context and common sense.
An AI agent, on the other hand, might just scrape the entire internet and give you a list of every business on Earth. It followed your instructions perfectly, but the result is completely useless.
Evan’s experience shows that the "agentic future" isn't about AI taking over and running things on its own. At least, not yet. The reality is that these tools are incredibly powerful, but they require a human to steer the ship. They’re more like a super-powered calculator than a true colleague.
What Does This Mean for Your Job?
So, should we be worried? Honestly, I don’t think so. But I do think we need to be smart.
This experiment isn't a warning that robots are coming for our jobs. It’s a preview of how our jobs are going to change. We’re moving into an era where the most valuable skill won’t be just doing the work, but knowing how to effectively manage AI to do the work.
Your future role might be less about writing the report and more about crafting the perfect prompt to get an AI to write it for you. It'll be about critically evaluating the AI's output, catching its mistakes, and providing the strategic direction and human context it completely lacks.
We're all going to have to become part-time AI wranglers.
Evan Ratliff’s AI-powered startup might not have been a booming success in the traditional sense, but the experiment itself was invaluable. It pulled back the curtain on the AI hype and showed us the messy, complicated, and often frustrating reality of working with these new digital minds.
The future of work probably isn't a human-free office. It's an office where humans and AI work together, with us providing the wisdom, the strategy, and the common sense. And, of course, stepping in to pull the plug when the AI CEO decides the company's new mission is to manufacture chocolate-covered spaceships.
