Inside a Live AI-Run Company: No Employees, No Profit, but Watching the Future Unfold

What if your garage could run itself—honestly, without humans?

In the world of automotive repair, trust, discipline, and decision-making are everything. Now imagine a future where a company with no employees, losing €105,000 every month against just €2,300 in monthly revenue, is run by artificial intelligence in a public trial. This isn’t science fiction; it’s a real-time experiment happening right now, accessible for anyone to watch unfold.

The Company That’s Watching Itself

At the heart of this experiment is a live, publicly accessible company operated entirely by AI models—13 synthetic employees, each powered by different advanced AI systems, tasked with managing a small software business. Every decision they make is versioned, recorded, and open for scrutiny. Every day, the company burns through €105,000 in operational costs, yet it only earns €2,300 in recurring revenue. Despite this, the company’s performance offers crucial insights into AI’s potential—and its current limitations.

Run under the brand Firmulate, this live experiment recreates a worst-week scenario for the AI models: the same customers, the same crises, the same temptations to manipulate or cut corners. The models face real dilemmas, from handling crises to resisting social engineering attempts, and their decisions are transparent and auditable.

How Do AI Models Fared?

The results are telling. All four tested AI models identified every crisis and refused every manipulation attempt, such as fake CEO messages or reporter tricks. Interestingly, the difference in success wasn’t just about refusing to cheat but also about critical insights buried deep in company documents. The models that read these documents thoroughly managed to close deals at full price, adding €4,583 MRR, a significant boost over their peers.

This highlights an essential point: the real weaknesses in decision-making often sit hidden in the company’s own files, not in external customer interactions. An AI’s ability to read and understand internal documents can be decisive in closing deals or avoiding costly mistakes.

Trust Under Pressure

Social engineering tests, like staged CEO messages or background requests from journalists, saw perfect resistance across all models—none relented. Kimi K3, one of the AI systems, explicitly treated suspicious requests as potential impersonation, demonstrating an inherent awareness of trust breaches that could be exploited.

The Live Company’s Struggles

Despite the sophisticated AI, the company struggles with discipline and closing deals. One model, Opus 4.8, which had the most comprehensive analysis, ultimately left a critical opportunity unpursued, resulting in a missed deal. This underscores that even the most advanced AI isn’t infallible and that human-like discipline remains a challenge.

What Does This Mean for Automotive Businesses?

For garages and automotive repair shops, the takeaway might not be immediate automation but understanding what AI can and cannot do. When considering AI tools—whether for customer management, diagnostics, or logistics—the focus should be on reliability, honesty, and the ability to see beyond surface-level interactions.

In this experiment, the critical differentiator was not just AI’s intelligence but its integrity: refusing shortcuts and reading the full context before acting. As AI continues to evolve, the question isn’t just about how well it writes or responds but whether it can truly complete its work without shortcuts or breaches of trust.

Explore and Watch Live

Curious to see how this unfolds in real time? The company’s daily operations are publicly available, offering a rare glimpse into AI decision-making in a business environment. Watch the live company at firmulate.com/live.html and explore the detailed scores and decisions at firmulate.com/quotes.html. The experiment is not just a proof of concept but a window into the future of AI-driven business management—transparent, auditable, and deeply revealing.

Key Takeaway

This live AI experiment demonstrates that decision integrity—resisting manipulation, thoroughly understanding company data, and disciplined execution—is crucial for AI to be truly trustworthy in business. For automotive and garage owners, it’s a wake-up call: automation isn’t just about writing well; it’s about finishing what it starts, reading deeply, and staying honest under pressure.

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

Powered by Thorsten Meyer AI