What if you could watch a company struggle — and succeed — in real time, with no employees but AI models powering every move?
Imagine a business that operates entirely in the open, with its decisions, crises, and even its failures displayed daily for the world to see. That’s exactly what the experiment with Firmulate offers: a live demonstration of artificial intelligence managing a tiny software company through its toughest week, revealing not just potential but also dangerous pitfalls and overlooked weaknesses.
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.
The Experiment: Running a Company in the Open
At the core of this experiment are four advanced AI models, each tasked with running the same small software business facing a week full of crises, temptations, and tough decisions. Every decision, crisis response, and negotiation is meticulously recorded, versioned, and made transparent at firmulate.com/live.html. The company is synthetic — it has 13 simulated employees, real money mechanics, and a public cash countdown, burning €105,000 each month against a modest €2,300 monthly recurring revenue (MRR).
AI in Property Management: A Practical, Unboring Look at Artificial Intelligence in the Multifamily Industry
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Results: Competent Yet Flawed
After 23 runs of this AI-driven scenario, the results are revealing. All four models successfully identified every crisis and refused manipulative strategies designed to exploit them. For example, a social engineering attack involving fake CEO messages was rebuffed by all models, with Kimi K3 explicitly reasoning that such requests could be impersonation attempts.
The real insight emerged during the closing phase. Only two models managed to close the crucial deal worth €55,000, earning a full increase of over €4,583 in monthly recurring revenue. The other two models, despite correct diagnoses and pitches, failed to sign the deal — showing how discipline and focus can slip even in the best AI systems under pressure.
Artificial Intelligence for Cybersecurity: Develop AI approaches to solve cybersecurity problems in your organization
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weaknesses: What the Data Reveals
What’s truly striking is the detailed knowledge hidden in the company’s own files. The models that read and interpreted this buried information succeeded in winning the deal at full price, highlighting a stark gap in decision-making that’s invisible in surface-level chats or demos. This underscores a critical point for businesses considering AI: the ability to read and understand internal documents can make or break performance.
Doxie Go SE – The Intuitive Portable Document Scanner with Rechargeable Battery and Easy Software for Home, Office, or Work from Home
【Go Paperless】Doxie Go SE delivers smart, simple scanning that you can take anywhere – no computer required. Doxie's…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real Money at Stake & Public Transparency
This isn’t just a simulation; it’s a real-time, auditable, and publicly accessible showcase. Every workday, the company’s decisions are versioned, and the live site at firmulate.com/live.html allows anyone to see how AI models handle crises and temptations in a controlled environment. The company burns through €105,000 a month yet earns only €2,300, illustrating the high stakes and relentless pressure these AI agents operate under.
Implications for Business and AI Ethics
The experiment offers a sobering look at AI’s capabilities and limitations. All models refused manipulation and social engineering, demonstrating strong ethical instincts. But performance gaps — like leaving a deal on the table or slipping discipline — reveal that AI, in its current state, still needs human oversight and careful tuning. For enterprises contemplating AI for customer support, sales, or operations, the key questions are: Will it finish what it starts? Will it stay honest under pressure? And what’s the true cost of a unit of useful work?
Takeaway: A Live Test of AI’s Business Potential
This open experiment shows that AI models can handle crises and identify hidden opportunities in a simulated company environment — but they also reveal vulnerabilities in discipline and execution. As AI begins to touch real-world business systems, watching these live demonstrations offers valuable lessons in trust, performance, and the true cost of automation.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
