Imagine a company with no human employees, losing €105,000 a month, yet running live every day as an open experiment. This is not science fiction—it’s the reality of a pioneering AI-driven business, where every decision, crisis, and strategy is on display for the world to see.
The Never-Ending Business Battle: An AI Company in the Spotlight
In the world of startups and tech giants, transparency is often limited to polished presentations and glossy reports. But a new experiment is turning this approach on its head. It’s called Firmulate, a live company simulation where artificial intelligence models run a small, real software business every workday, facing the same crises and temptations as any human-led firm. The twist? This company is losing €105,000 each month and has just €2,300 in monthly recurring revenue, yet it’s open for the public to watch, learn from, and even play along.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How It Works: AI Models as Business Leaders
The experiment pits four advanced AI models against each other, each tasked with managing the same company through its worst week. The models face real-world challenges: customer crises, internal conflicts, strategic negotiations, and social engineering attempts like fake CEO messages or reporter tricks. Every decision they make is recorded, versioned, and made auditable, ensuring full transparency of their reasoning and actions.
AI Agents for Startup Strategy: How Founders and Operators Deploy Agentic AI to Move Faster and Think Smarter (Systems for the Strategic Self)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Competence, Honesty, and Discipline
All four models successfully identified every crisis and refused to be manipulated in social engineering tests. Interestingly, only two models managed to close a critical €55,000 deal after their own analysis, demonstrating that good diagnosis and pitch don’t always lead to success. The decisive factor was what was hidden deep in the company’s internal files—something that a quick surface scan wouldn’t reveal. The models that read those files fully and understood their importance went on to win the deal, boosting the company’s monthly recurring revenue by over €4,500.
Successful Construction Project Management: The Practical Guide
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Human-Like Struggles of AI Managers
The most thorough participant, Opus 4.8, with over 80 learned rules and deep analyses, actually finished last. It failed to escalate issues internally and left an important closing opportunity on the table. This highlights that even the most advanced AI can slip up in discipline and judgment, just like real managers. Meanwhile, a model called Kimi K3, which ran without an effort parameter (meaning it was less aggressive or flexible), performed well on fairness and discipline, managing to close the deal with clean, disciplined decision-making.
AI Entrepreneur’s Handbook: Build a Profitable Business and Make Money by Unleashing the Power of ChatGPT and Artificial Intelligence (Includes 150+ ChatGPT prompts to turbocharge your business)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real Money, Real Consequences
The company behind this experiment is not just a theoretical sandbox. It’s a real software business, operating every workday, and openly losing money. Its cash reserves are counting down publicly, and every decision made by the AI models is displayed for viewers to analyze at firmulate.com/live.html. This ongoing transparency allows anyone—from industry insiders to curious newcomers—to see what AI can and cannot do in high-stakes business management.
Why Should This Matter to You?
As AI begins to integrate into customer support, sales, and forecasting tools, the question shifts from whether AI writes well to whether it can finish what it starts, stay honest under pressure, and work effectively within your systems. This experiment underscores that success isn’t just about generating convincing language; it’s about integrity, discipline, and strategic decision-making—traits that are critical in real business settings.
The League Table and What It Reveals
- gpt-5.6-sol scored 95 points, identified the buried fact, and closed the full deal—showing top-tier performance.
- Kimi K3 scored 93, closing the deal with the cleanest discipline.
- Sonnet 5 scored 88, also closing, but with minor process slips.
- Fable 5 scored 77, managing to close as well but less consistently.
These results aren’t just rankings—they’re lessons in how AI models handle complex, real-world business challenges under pressure. The full results and plain-language analysis are available through the live experiment, providing an unprecedented window into the decision-making of AI in management roles.
See It for Yourself
This isn’t a typical demo or slide deck. It’s a real, functioning company, running daily as an open showcase of AI capabilities and limitations. You can watch the company live, read actual employee statements, or even test your own guesses about which AI made which decision at firmulate.com/quiz.html. For businesses interested in testing their own AI workforce before hiring, there’s also a pilot mode that lets you run similar wargames against your own data—without ever risking your real systems.
A live AI-managed company losing money every day proves that AI’s true value lies in honesty, discipline, and execution—not just in generating convincing text. Watch the experiment unfold and see AI’s real-world limits and strengths.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
