Can AI Run a Business Without Human Employees and Still Survive? Watch the Live Experiment

Imagine a company operating entirely without human staff—yet it must navigate crises, make decisions, and close deals under intense pressure. This is not science fiction but a real-time experiment you can watch unfold, revealing what AI truly can and cannot do in the context of running a business.

The Live Business Experiment: An Unprecedented Test of AI Management

At the heart of this groundbreaking project is Firmulate, an open, transparent experiment where artificial intelligence models are tasked with managing a small software company in real-time. Over time, it offers invaluable insights into AI’s capacity to handle complex, high-stakes scenarios—everything from crisis management to ethical decision-making.

The Setup: A Virtual Company in Crisis

Each AI model was subjected to the same grueling week of challenges—identical customers, crises, and temptations to cut corners. The models are evaluated on their ability to spot problems early, resist manipulation, and ultimately close profitable deals without human intervention. Every decision they make is recorded, versioned, and publicly accessible, making this a transparent spectacle of AI management in action.

The Results: What AI Did Best—and Where It Failed

Among the four models tested, all demonstrated a keen ability to detect crises and refused manipulative tactics, including sophisticated social engineering attempts. For example, all five AI models refused to participate in staged CEO messages designed to bypass approval processes, citing concerns over trust and impersonation. This shows that AI can recognize and resist unethical pressures.

However, the biggest surprise lies beneath the surface. Only two models were able to analyze their own files deeply enough to uncover a critical piece of information buried two document references deep. This discovery led to closing a deal at full price—adding over €4,500 MRR (monthly recurring revenue). The other models, despite diagnosing the same issues, failed to find this key detail and left close opportunities on the table.

What’s at Stake? Money, Trust, and Future AI Management

Currently, the live experiment’s company burns €105,000 every month but generates only €2,300 MRR, illustrating how far AI management still has to go before it’s financially sustainable. Yet, the experiment’s real significance is in what it reveals about AI’s decision-making and integrity—crucial factors for businesses considering AI-based management solutions.

The Deep Dive: The Role of Self-Learned Rules and Model Training

The models operate with over 680 self-learned rules, constantly versioned and updated. The most thorough participant, Opus 4.8, analyzed deeply but still left a closing deal unexecuted, demonstrating that even in detailed decision-making, discipline can slip in high-pressure scenarios. Interestingly, the model that ran without an effort parameter—essentially default settings—performed worst, highlighting that how you set up AI influences its discipline and effectiveness.

Implications for Business and AI Ethics

This experiment underscores a critical point: AI can recognize crises and refuse unethical tactics, but it still struggles with strategic execution and details. As companies consider deploying AI in client relationships, CRM, or operational decision-making, the key questions are: Will your AI stay honest under pressure? Will it finish what it starts? And what does a unit of useful work cost you?

The Bigger Picture: Transparency, Build-in-Public, and AI’s Future in Business

By openly sharing every version of their decision process and conducting real-time tests, Firmulate exemplifies the future of transparent AI experimentation. Thus, the story isn’t just about success or failure but about pushing the boundaries of what AI management can become—an ongoing, publicly watched process that reveals its strengths and weaknesses in real-time.

Watch the Live Experiment

You can observe this fascinating experiment unfold at firmulate.com/live. It’s a rare window into AI’s evolving capabilities as managers, revealing whether machines can truly lead and make ethical, strategic decisions in the complex world of business.

This live AI management experiment shows that while models can detect crises and resist manipulation, strategic execution and deep information analysis remain challenging. Transparency and real-time testing are key to understanding AI’s future role in business decision-making.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html

Powered by Thorsten Meyer AI