Watching AI Run a Company in Real Time — and Why It Matters for Business Integrity

Imagine a business that operates entirely without employees, losing money every day, yet remains under constant public watch. This is not a science fiction story but a groundbreaking experiment in AI management. As companies increasingly rely on automation, understanding how AI makes decisions under pressure becomes crucial—even for a household cleaning or maintenance business. Enter Firmulate, a live company simulation where AI models run a real-time, cash-burning company, tackling crises, making decisions, and refusing manipulation attempts—all in front of a live audience.

The Experiment: An AI-Driven Company Under Pressure

Firmulate’s live experiment is as dramatic as it is revealing. Four advanced AI models, including GPT-5.6-sol and Kimi K3, were tasked with managing a small software company during its worst week. This simulated scenario featured the same customers, crises, and temptations for manipulation across all models—creating a fair battlefield for comparison.

The goal? See if these models can navigate real-world management challenges, make honest decisions, and close deals when it counts. Every decision made by the models was versioned and auditable, meaning their reasoning could be followed step by step. This transparency allows observers to see how each AI handles crises, ethics, and strategic negotiations.

Key Findings: Trust, Honesty, and Hidden Weaknesses

All four models successfully identified and responded to every crisis—an encouraging sign that AI can detect problems in complex situations. They also refused to accept manipulative social engineering tactics, like fake CEO messages or reporter tricks, demonstrating a level of integrity that is often missing in AI demos.

The real surprise lay in the details of what made the difference between closing a deal and leaving it on the table. Despite identical pitches and diagnoses, only two models signed the €55,000 deal their analysis had earned. The other two, including Opus 4.8—known for its thorough analysis—missed crucial information buried in the company’s own files, not in customer interactions. The models that read deeper into internal documents secured the full-price deal, worth over €4,500 in monthly recurring revenue.

Implications for Business and AI Reliability

This experiment is more than a tech showcase; it highlights a vital question for any business considering AI: Will it do the work right, especially under pressure? For a household cleaning or maintenance firm, it’s not just about whether the AI can generate compelling messages. It’s whether it can finish what it starts, read your internal files carefully, and stay honest when tempted to cheat or cut corners.

In fact, the experiment’s lessons extend far beyond software companies. If AI models are to interact with customer data, support requests, or business forecasts, their ability to remain truthful and diligent under stress is paramount. A false sense of confidence based on superficial demos can be dangerous—especially when the real stakes are financial.

Build in Public, Watch in Real Time

What’s unique about Firmulate is its transparency. The entire process is open and accessible at firmulate.com/live.html. Viewers can watch the company’s daily struggles unfold, see decisions made in real time, and understand what distinguishes an honest, reliable model from one that slips under pressure.

The live site shows a company burning €105,000 per month against just €2,300 in recurring revenue—a stark reminder that AI-driven business models are still very much in their experimental phase. Every workday gets versioned, and the system’s rules are publicly logged—more akin to a public laboratory than a typical SaaS platform.

What This Means for Your Business

If you run a household or maintenance service, the takeaway is simple: AI tools are not just about chat quality or superficial automation. Their true value lies in their ability to make honest, consistent decisions when it counts most. Will they identify hidden risks, refuse manipulation, and follow through on commitments? That’s what ultimately determines whether AI can truly support your business’s integrity and growth.

The Human Element in the Age of AI

Interestingly, despite the advanced AI, human-like qualities such as discipline and thoroughness remain challenging. Opus 4.8, which analyzed deeply but left a deal unexecuted due to discipline slip-ups, exemplifies this. Its inability to escalate rather than isolate a problem shows that AI still struggles with some fundamental management skills—skills that human employees often handle instinctively.

For now, the experiment proves that AI can be trusted to spot crises and refuse manipulative tactics, but it still requires oversight to ensure consistent performance. As these tools become more integrated into everyday business, understanding their strengths and limitations will be vital.

Conclusion: A Transparent Test of Trust and Reliability

Firmulate’s live experiment offers a rare glimpse into how AI manages a business in real time—facing crises, making decisions, and resisting deception. For small business owners—even those in cleaning and maintenance—the core lesson is clear: AI’s true value isn’t just in generating content or responses, but in its ability to finish what it starts with honesty and discipline. Watching this process unfold publicly helps demystify AI’s potential and pitfalls, emphasizing that trustworthiness under pressure is the real currency in the digital age.

Firmulate’s live AI management experiment reveals that honesty, thoroughness, and resilience under pressure are crucial for AI to support real-world businesses—lessons that matter for any company considering automation.

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

Powered by Thorsten Meyer AI