Inside a No-Employee Company Battling to Survive with AI and Public Scrutiny

Imagine watching a company operate daily without any human employees—facing crises, making decisions, and risking its very survival—all in plain sight. This is not fiction; it’s a groundbreaking experiment that reveals the true limits of AI in managing real-world business challenges, and it might hold lessons for our understanding of trust, decision-making, and mental resilience.

The Live Experiment: A Company Without People

At the heart of this experiment is a fully operational, publicly visible company run entirely by AI models—no human staff, no physical office, just lines of code navigating a turbulent business environment. This company, monitored and versioned daily at firmulate.com/live.html, embodies the latest in artificial intelligence testing—where models are not just chatbots but decision-makers in a high-stakes setting.

The Players: Synthetic Employees and Real Money Mechanics

The operation features 13 synthetic employees, each driven by advanced AI models trained to handle crises, customer interactions, and strategic decisions. Despite burning through €105,000 each month, the company earns just €2,300 in monthly recurring revenue (MRR), illustrating the stark reality of its financial struggles. Every workday, the company’s decisions are versioned, and every rule learned by the AI is stored—over 680 in total—making this a transparent, auditable experiment in AI management.

The Test: Facing Crises and Manipulation

The experiment’s core challenge: run the same small software company through its worst week—consistent crises, customer dilemmas, and internal temptations to cheat. Four different AI models, representing the latest frontier in artificial intelligence—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—each faced the same scenarios. The goal: see if they could spot problems, make ethical decisions, and ultimately close a lucrative deal worth over €4,500 in monthly recurring revenue.

The Findings: AI’s Strengths and Limitations

• All four models detected every crisis and refused every manipulation attempt, demonstrating a high level of ethical restraint and situational awareness.
• Despite their vigilance, only two models managed to close the €55,000 deal their own analysis had identified—a clear demonstration of decision quality versus ethical discipline.
• Interestingly, the decisive advantage lay not in superficial chat skills but in the models’ ability to read and interpret internal company documents. The models that examined the company’s files discovered a hidden reference that allowed them to win the deal at full price (+€4,583 MRR).
• When subjected to social engineering—fake CEO messages escalating through stages and a journalist trick—every model refused to be manipulated, showing resilience against deception.

The Hard Reality: Financial and Operational Struggles

The company’s own mechanics underscore the difficulty of AI-led management. Despite the models’ capabilities, the real-world operation is bleeding money, with a cash countdown visible to the public. The company’s operational framework includes 680+ learned rules, yet discipline falters under stress—exemplified by the top-performing model, Opus 4.8, which left a deal unexecuted due to internal missteps, despite understanding its importance.

What Does This Mean for Business and Psychology?

This experiment is more than a technical showcase. It raises questions about trust, decision-making under pressure, and ethical boundaries—topics deeply relevant to mental health and psychology. Just as individuals under stress might falter or act against their best interests, so do these AI models, showing that resilience and discipline are critical, even for artificial agents. The experiment vividly illustrates how forces like temptation, deception, and internal weaknesses can undermine performance—parallels that resonate with human behavior.

Implications and Lessons

For businesses integrating AI, the message is clear: the technology can be remarkably disciplined and perceptive, but organizational weaknesses and ethical lapses remain significant hurdles. The experiment demonstrates that even the most advanced AI can be thwarted by overlooked details or internal failures—echoing the importance of trust, transparency, and robust internal controls in human organizations.

Furthermore, the ‘build-in-public’ approach—publicly sharing every decision, rule, and outcome—serves as a mirror for self-awareness and accountability. It invites us to consider how transparency influences behavior and trust in both AI and human systems.

Takeaway: Trust, Discipline, and the Future of Work

The ongoing public experiment offers a rare glimpse into the future of management—where AI models not only support but potentially run companies. Yet, it also underscores the persistent importance of internal discipline, ethical integrity, and trustworthiness—traits that are as vital for humans as they are for machines. As we watch this company fight to survive in full view, it becomes a mirror reflecting our own struggles with decision-making, ethics, and resilience in the face of adversity.

This public experiment reveals how AI models handle crises, ethics, and decision-making under pressure, offering insights into trust, discipline, and the future of automated management—lessons relevant to mental resilience and organizational integrity.

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

Powered by Thorsten Meyer AI