The crazy question that started it
After years of businesses and teams of 20 to 30 people, I wanted a radical experiment in the opposite direction. How far could I build a real, growing business without permanent employees? I didn't mean a lifestyle freelance business. I wanted products, customers, revenue and a high margin.
AI made the question bigger. If research, content, analysis, support and many handoffs no longer need a person every time, how far can one person take a business? Put provocatively: 'One Person, One Billion Business?'
Talking about it wasn't enough for me. I wanted to live it, with all the bad decisions, limits on capacity and system problems. The products and views I stand for today grew out of that experiment.
The three stages: Tool, workflow, AI employee
There are several steps between a helpful chat answer and an AI employee working independently. Skip them and you can quickly spend more time correcting mistakes than the original task took. That's why I distinguish these three stages.
1. Chat-based AI
You give the chat a task and get an answer. That helps. But you still have to start the next task yourself.
2. Automated workflows
Recurring steps are connected through tools and triggers. The process runs in a standard way but remains rigid when exceptions arise.
3. AI employees
The AI has a clear task, the information it needs and access to suitable tools. It acts when something happens and reports its result.
What already works surprisingly well
AI can research leads, prepare meetings, structure content, help with customer questions, analyze product feedback and bring internal information together. In clearly defined roles, that becomes more than a time saving: work gets done without a person having to start every step.
AI works best on tasks involving lots of language, recurring context and a result you can clearly check. Be careful when goals are vague, AI communicates externally without review or a mistake could have serious consequences.
I need to see what the AI did and where it's stuck. Handing work over and never wanting to hear about it again was already a bad idea with people. With AI, mistakes can simply repeat faster.
- Sales: Research, qualification, briefings and follow-up preparation.
- Marketing: Source research, variants, editorial workflow and distribution.
- Customer success: Context-aware help, encouraging use and escalation.
- Operations: Summaries, handoffs, documentation and quality checks.
- Product: Grouping feedback, tests, specifications and ongoing improvement.
Why the experiment didn't end with doing everything alone
The AI apps and demand for implementation helped the business grow. But in the end, there was still me. I had to decide, maintain relationships and judge where we should go next. AI could do a lot. I still became the bottleneck.
So I wanted people in the business again who could lead their own area. People I could think ideas through with, who would notice when something was wrong. AI takes a lot of work off their hands. That makes their experience and judgment more valuable. That was the correction to my original idea.
Today, we work with four people and more than 50 autonomous AI employees. That ratio isn't a trophy. It's a practical indication of how differently a business can be structured when growth doesn't automatically create more hierarchy.
How the experiment led to CoachWunder
Much of my work revolved around making expertise personally useful instead of leaving it sitting in content. Customers needed answers, guidance and help doing the work at the moment the situation arose. The next call was often too late.
CoachWunder emerged as a platform for personal AI products. Experts upload their knowledge, method and context. In a few hours, that can become a first AI coach or working product that interacts with real users.
For me, this isn't a theoretical AI use case. Years of digital products have shown me that customers don't need another mountain of content. They need help that moves them closer to their result.
Why ConnectWunder had to follow CoachWunder
Each new AI employee created a new coordination problem. Context was scattered across chats, meetings, projects, customer data and people's heads. A regular messenger or traditional project tool doesn't automatically understand those connections.
When four people work with more than 50 AI systems, opening another chat for every process isn't enough. People and AI need a shared workplace connecting customers, tasks, decisions, knowledge and current events.
That's why we built ConnectWunder. The latest customer decision should live where the people and AI continue their work. Otherwise, someone spends the day carrying information from one chat to another. That's exactly the work we wanted to get rid of.
Freedom, revenue and margin belong together
I never wanted to just squeeze more work into less time. A business that only runs while the founder pushes everything forward isn't freedom to me. But thinking small isn't the answer either. The business needs to genuinely help customers, stay profitable and grow without hiring another person for every new order.
AI can lower costs and get more work done. That only stays profitable if the product has clear boundaries, quality is visible and the team doesn't have to rescue every exception by hand.
So I don't want a business empty of people. I want a business where people do their best work, AI handles the dull routine tasks and growth doesn't automatically add more management layers.
Seven lessons from the ongoing experiment
The experiment isn't finished. That's why it remains the basis of my work. I share decisions from products and businesses operating under real pressure today. I don't have a perfect blueprint from the future.
- Start with real tasks, not an abstract agent strategy.
- An AI employee needs to know what to do, what it's allowed to do and who gets the result.
- Hand work over and watch the results. Stepping in is part of leading.
- Don't automate a process if you can't clearly describe its result.
- AI makes strong people more valuable, not redundant.
- If every chat knows something different, you have to bring it all together yourself again.
- Check whether customers make progress and money is left over. An impressive demo tells you neither.
Frequently asked questions
Questions about this article
Does Sebastian Maier really have no employees?
That was the original idea. Today, we work with four people and more than 50 AI employees. The experiment taught me how much stronger the right people can be with AI. That's why I changed my mind.
What is an AI employee?
An AI employee handles a specific task independently. It has the information and tools it needs and reports back on what's done or where it needs a decision.
Can a one-person business make millions in revenue with AI?
AI greatly changes how much one person can get done. Revenue still depends on the market, offer, sales, quality and responsibility. Technology alone doesn't guarantee a business model.
Why do we need ConnectWunder?
Because people and multiple AI systems need shared context about customers, tasks, decisions and knowledge. Traditional isolated chats don't solve that coordination problem.