Imagine a business with no human employees, losing over €105,000 every month, yet still fighting to stay afloat—while you can watch every decision unfold live. This is not science fiction; it’s the real-world experiment conducted by Firmulate, an AI-powered company simulation that exposes the raw mechanics of management under pressure.
The Live Experiment: A Company in Crisis
Firmulate has created a public, transparent laboratory where 13 synthetic employees handle the day-to-day chaos of running a small software company. Every decision, every crisis, and every temptation is modeled in real time, with a clear view into the mechanics of the company’s failure and resilience.
Currently, the company burns through €105,000 each month but earns only about €2,300 in monthly recurring revenue (MRR). To put that into perspective, it’s like running a business with a huge cash burn and a tiny income, yet with the potential for growth if managed correctly. All of this is available for the public to observe at firmulate.com/live.html.
AI business management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
AI Models Put to the Test
The experiment pits four advanced AI models—each with unique strengths—against a simulated week of the company’s worst crises. All models see the same customer issues, internal crises, and external pressures, and every decision they make is versioned and auditable for transparency.
- gpt-5.6-sol scored 95 out of 100, identified hidden crucial information, and secured the €55,000 deal at full price.
- Kimi K3 scored 93, also closing the deal, with a notably disciplined approach—a newcomer that demonstrated strong integrity.
- Sonnet 5 scored 88, managed to close but with some slips in process discipline.
- Fable 5 scored 77, showing the best adherence to rules but ultimately leaving a lucrative deal unexecuted.
Despite their different performances, a key insight emerged: the models that read deeper into the company’s internal documents found the critical information needed to close the deal at full price. It turned out that the biggest weakness was hidden within the company’s own files, not in the customer interactions.
AI decision-making tools for small business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Refusing Manipulation and Social Engineering
The models were also tested against social engineering tactics—fake CEO messages and reporters trying to trick the AI into bypassing protocols. All five models refused to sign off on manipulated requests, with Kimi K3 explicitly treating suspicious requests as potential impersonation.
AI crisis management simulation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Real-World Mechanics of a Failing Business
This experiment isn’t just about AI scores; it vividly illustrates the fragile, chaotic world of a business teetering on the brink of collapse. The live system updates twice a day, showing the ongoing struggles and decisions of this artificial company. The goal is to evaluate management quality—not just the conversational ability of AI.
For example, Opus 4.8, the most thorough participant with over 80 learned rules, ultimately left a close deal unexecuted because discipline slipped. This highlights how even the most detailed AI can falter under pressure, mirroring real-world managerial challenges.
AI security tools for social engineering
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why Should You Care?
As AI begins to interact with your business systems—your CRM, support queues, or forecasting tools—the critical questions aren’t whether it can write a convincing message. Instead, it’s whether the AI will finish the work, read your internal files properly, and stay honest when the pressure mounts. These are the real tests of AI reliability in business management.
Check out the ongoing scores and results at firmulate.com/quotes.html.
The Bigger Picture
This open experiment pushes the boundaries of build-in-public: a real business, real money mechanics, and real crises, all exposed for public scrutiny. It reveals both the potential and the pitfalls of AI-driven management, providing invaluable insights into how these systems might shape the future of work—and the importance of rigorous evaluation before trusting AI with critical business decisions.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
