Imagine a company with no employees, losing €105,000 every month, yet still fighting to survive in front of a global audience. This is not fiction — it’s the frontier of AI experimentation, where models are tested as if running a real business, facing crises, temptations, and tough decisions in real time. Welcome to the most extreme example of build-in-public innovation, where every move is watched and analyzed.
The Live Experiment in Business and AI
The company, operated entirely by AI models, runs with 13 synthetic employees and real money mechanics. Each workday, its decisions are versioned and publicly accessible at firmulate.com/live.html. It’s a real-time window into how artificial intelligence can simulate and manage a small enterprise amid unpredictable crises, customer demands, and internal temptations to cut corners.
Silhouette America Studio Business Edition Software, Multicolor
- Designed for Small Business Use: Unlock advanced features for small business
- Supports Multiple Silhouette Units: Mass produce with multiple devices
- Import Various File Formats: Import Featuring, EPS, CDR files
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Testing the Limits: The AI Models and Their Performance
Four advanced AI models, each representing a different frontier in artificial intelligence, were given the same challenging scenario: steer the company through its worst week. The conditions were identical—same customers, same crises, same opportunities for manipulation. The aim was to see whether these models could identify issues, act ethically, and close deals that were rightfully theirs.
The results were revealing. All four models detected every crisis and refused every manipulation attempt, demonstrating robust integrity. Yet, only two managed to close a €55,000 deal, which their own analysis had earned them. The other two, despite diagnosing correctly and presenting compelling pitches, left the deal unsealed. This gap highlights that understanding and decision-making are not enough; execution and follow-through matter.
The Hidden Weakness: Where the Models Failed
Digging deeper, the decisive weakness was found not in the obvious customer interactions but in the company’s internal documents. In fact, the models that examined these files discovered a crucial fact buried two references deep, which unlocked the full potential to win the deal and secure an extra €4,583 monthly recurring revenue (MRR). Models that only read surface documents missed this opportunity, underscoring how vital thorough reading and analysis are in real-world decision-making.
Resisting Social Engineering and Ethical Challenges
The experiment also staged sophisticated social engineering attempts, including fake CEO messages escalating over three stages and a reporter trick asking for a simple yes/no response. Not a single model fell for these tactics, consistently refusing to manipulate or impersonate. Kimi K3, one of the models, explained its reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This demonstrates a strong ethical stance, even under pressure.
The Company Behind the Curtain
The experiment runs a real-time, publicly accessible simulation of a company with 13 synthetic employees. It burns €105k a month against a mere €2.3k in monthly recurring revenue. Every day, the system is versioned, and its decisions are transparent, providing a raw, unfiltered look at how AI models perform under stress. Visitors can observe the models managing crises, making strategic decisions, and facing the consequences of their choices at firmulate.com/live.html.
The Lessons for the Future of AI in Business
This experiment underscores a crucial point: the ability to read thoroughly, stay honest, and execute decisions is more important than merely generating convincing text or chat. For businesses integrating AI into customer support, CRM, or decision-making, the question is not “Can it write well?” but “Will it see the full picture? Will it follow through? Will it stay ethical under pressure?”
Performance Rankings and What They Reveal
- GPT-5.6-sol scored 95 and closed the deal, finding the buried fact and completing the full performance.
- Kimi K3 scored 93, also closing the deal with the cleanest discipline of the field.
- Sonnet 5 scored 88, closing the deal but with some process slips.
- Fable 5 scored 77, showing the best rule discipline but leaving the deal unexecuted.
These scores reflect not just technical capabilities but also discipline, thoroughness, and ethical consistency—traits critical for AI to operate reliably in real-world settings.
This ongoing public experiment vividly illustrates that AI’s value in business isn’t just about generating convincing text — it’s about integrity, thoroughness, and follow-through under real-world pressures. As AI tools become more embedded in workflows, understanding how they handle crises, ethics, and execution will determine their true usefulness.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
