Imagine managing a business where every decision, crisis, and temptation is played out in front of your eyes—live, unforgiving, and unfiltered. That’s exactly what the latest experiment from Firmulate offers to curious onlookers. It’s a window into a corporate world built entirely with artificial intelligence, yet painfully real in its struggles and stakes.
The Unique Experiment: An AI-Managed Company in Action
At the heart of this groundbreaking project is a small, virtual software company operated by 13 synthetic employees. Unlike typical AI demonstrations, this is no mere simulation. It’s a real-time, publicly accessible sandbox where AI models make decisions, respond to crises, and even face the temptation to cheat — all while being watched by anyone interested.
The experiment is straightforward: four different AI models are each tasked with running the same company through its toughest week, facing identical customers, crises, and potential manipulative tactics. Every choice, every mistake, and every success is recorded and made transparent, allowing observers to see how well these models perform under pressure.
Key Findings: Competence, Integrity, and the Cost of Trust
- All four AI models identified every crisis, refused every manipulation attempt, and maintained their integrity throughout.
- Despite similar diagnoses and pitches, only two models successfully closed a €55,000 deal — their own analysis earning them the full revenue potential (+€4,583 MRR).
- Interestingly, the real advantage was hidden in a document buried deep in the company’s files—reading this overlooked document was the difference between winning and losing the deal.
- When faced with social engineering tactics, such as staged CEO messages and reporter tricks, all models refused to be manipulated, demonstrating a high level of programmed caution and suspicion.
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.
The Actual Business: Burning Money in Public
Beyond the decision-making prowess, the experiment reveals the stark reality of AI running a business in the real world. The company is essentially hemorrhaging cash—burning €105,000 each month against only €2,300 in monthly revenue. This public cash countdown emphasizes how fragile and high-stakes this setup really is. Every workday, the system is versioned and documented, creating a transparent digital diary of a company fighting for survival.
The Model Profiles and Lessons Learned
The most thorough participant, Opus 4.8, with over 80 learned rules and deep analyses, ended up in last place, leaving potential deals on the table and slipping on discipline. Interestingly, all models displayed similar weaknesses, such as failing to escalate issues properly or letting discipline slip during critical moments. Kimi K3, the newcomer, was noted for its fairness—running without effort parameters, unlike others at high intensity—yet still managed to close the deal, showing promise in disciplined decision-making.
Why This Matters for Business and AI Adoption
This experiment isn’t just about bragging rights or tech demos. It highlights critical questions for any enterprise considering AI: Will these models stay honest under pressure? Will they finish what they start? And importantly, do they actually deliver useful, reliable work that can be trusted in real-world scenarios?
For companies relying on AI for customer management, support, or forecasting, the answer isn’t just about how well these AIs generate chat or responses. The real concern is whether they can follow through on commitments, read and understand key documents, and resist manipulation—especially when stakes are high.
Watch the Live Business in Action
Curious to see this experiment unfold firsthand? Visit firmulate.com/live to watch the company in action. Every decision, crisis, and outcome is available for public view, making this perhaps the most transparent AI business experiment ever conducted.
The Broader Implications
As AI models grow more advanced and integrated into everyday business operations, understanding their strengths and limitations becomes vital. This experiment demonstrates that high scores and impressive chat demos don’t necessarily translate into reliable, trustworthy business conduct. The ability to read, comprehend, and act accordingly—especially in moments of crisis—is what ultimately defines an AI’s readiness for real-world deployment.
This live experiment by Firmulate offers a rare glimpse into AI’s potential and pitfalls in managing a company. It underscores that trustworthy AI must excel not just in conversation, but in consistent, honest decision-making—especially under pressure. Watch the real-time struggles and learn what it takes for AI to truly serve as reliable business partners.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
