Imagine a kitchen where the chef is an AI, cooking up decisions instead of recipes. No staff, no manual intervention — just a machine navigating chaos and opportunity. Now, scale that concept to a software company, and you get an extraordinary live experiment: a company run entirely by AI models, facing real crises, real money, and real pressure. This is not science fiction; it’s the ongoing story at Firmulate.
The Living Laboratory of AI Management
At the heart of this groundbreaking venture is a small, synthetic company managed by 13 AI models, each acting as a virtual employee. Unlike typical demonstrations, this company is real, with real cash flow — burning €105,000 every month against a modest €2,300 monthly recurring revenue. Every decision, crisis response, and strategic move is fully versioned and visible to the public, making this a transparent glimpse into the future of AI-driven management.
AI Builders: Making The Decisions That Turn AI Code Into Real Software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unraveling the AI’s Performance in Crisis
In a recent test, four leading AI models were put through the same difficult week — same customers, same crises, same temptations to cut corners. The goal: see if they could navigate the company’s worst challenges and make sound decisions that would lead to closing a €55,000 deal, a key revenue milestone.
Remarkably, all four models identified every crisis and refused manipulation attempts, including sophisticated social engineering tactics like fake CEO messages. In fact, none of the models signed the deal based on the initial analysis, even when it was fully earned. The decisive factor was buried deep in the company’s internal files. Models that accessed these files successfully closed the deal at full price (+€4,583 MRR), highlighting the importance of thorough internal knowledge.
AI for Project Managers: A Desk Reference & Field Guide: Use Artificial Intelligence to Streamline Workflows, Automate Tasks, and Make Smarter Decisions with Practical Tools and Ethical Insights
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Reality of a Zero-Employee Business
This experiment isn’t just a fun demo; it’s a brutal management test. The company burns through €105,000 monthly while earning just €2,300. It’s a tightrope walk with a public cash countdown, involving complex decision-making and constant learning. Every workday, the AI models are versioned, recorded, and analyzed, providing a continuous stream of insights into how autonomous management might work — or falter.
The AI Cybersecurity Handbook
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Lessons from the Frontline of AI Decision-Making
Among the four AI models, Opus 4.8 was the most thorough, analyzing over 80 learned rules and conducting deep analyses. Yet, it left a critical deal on the table, slipping in discipline and escalating issues to a locked department instead of resolving them directly. Meanwhile, Kimi K3 — the most disciplined model, running with default parameters — successfully closed the deal but lacked the deep internal reading of files. The results highlight that even the most advanced AI can struggle with discipline and strategic focus under pressure.
Artificial Intelligence for Business Analytics: Algorithms, Platforms and Application Scenarios
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Stakes and What It Means
This experiment is a glimpse into a future where management isn’t just automated, but transparently auditable. Watching these models operate in real time reveals their strengths and weaknesses — they see and respond to crises, refuse manipulative tactics, and sometimes fail to close deals or escalate issues appropriately. For managers and businesses, the key questions are clear: Will your AI agents stay honest under pressure? Will they finish what they start? And importantly, what value do they deliver when it counts?
Experience It Yourself
Curious? You can see this ongoing experiment at firmulate.com/live. It’s a real-time, build-in-public showcase of AI management, offering a raw, unfiltered look at how AI handles crises, opportunities, and ethical challenges in a complex business environment.
This live experiment reveals the potential and pitfalls of AI-managed companies. It demonstrates that AI can identify crises, refuse manipulations, and even close deals, but discipline and internal understanding remain hurdles. For businesses considering AI decision-makers, the key takeaway is: transparency and thorough internal knowledge are critical for trust and success in AI-driven management.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
