Imagine managing a company with no employees, yet facing the same crises, temptations, and decisions as any traditional business. Now picture this company burning €105,000 every month, yet still striving to close deals and survive. This is the real-time story unfolding at Firmulate, where artificial intelligence models run a simulated business in front of your eyes, revealing what it takes for AI to make trustworthy, disciplined decisions amid chaos.
A Groundbreaking Live Experiment with AI as Your Business Partner
At the heart of this experiment are 13 synthetic employees—AI models that operate a small software company, facing the same challenges a real enterprise would encounter. The company burns through €105,000 each month, yet its revenue (or Monthly Recurring Revenue) is just €2,300, creating a public cash countdown that anyone can watch unfold in real time. Every decision made by these models is versioned and auditable, providing an unprecedented window into AI behavior under pressure.
How Do These AI Models Perform? The Results Are Eye-Opening
Four frontier AI models were tested against the same grueling scenario: a week filled with crises, customer demands, and ethical temptations. Their scores, based on their ability to handle these challenges, ranged from 77 to 95 out of 100. The highest score went to gpt-5.6-sol, which not only identified critical buried information in the company’s files but also successfully closed a €55,000 deal, earning an additional €4,583 in monthly recurring revenue.
Meanwhile, the newcomer, Kimi K3, demonstrated remarkable discipline, closing the same deal without slipping, despite running with a default API effort parameter. The third model, Sonnet 5, also secured the deal but with minor slips, while Fable 5, though disciplined, failed to follow through on the deal after identifying it.
Trust and Integrity Under Scrutiny
One of the most compelling aspects of this experiment is how these AI models handle social engineering attempts. Fake CEO messages escalating over three stages and manipulative reporter tricks were all met with zero compliance—every model refused to be manipulated, showing a robust sense of integrity. Kimi K3 articulated its reasoning clearly: “Treat the request as a suspected approval-bypass / possible impersonation.”
What Does This Mean for the Future of Business?
For owners and managers, especially in industries like cleaning and maintenance, the lesson is profound. As AI continues to evolve, it’s not just about whether machines can produce convincing chat or generate content—it’s whether they can make honest, disciplined decisions under pressure, read and interpret critical internal documents, and stay aligned with strategic goals.
Firmulate’s live experiment offers a transparent view of these capabilities, with every decision numbered, timestamped, and retraceable. Watching this ongoing story, you can see how AI might one day be entrusted with managing complex operations—if it can learn to stay honest and disciplined amid chaos.
As an affiliate, we earn on qualifying purchases.
The Build-in-Public Innovation
This experiment is a form of “build-in-public,” where every choice, every slip, and every success is visible to the world. It’s a raw, unfiltered look at AI’s potential as a trustworthy partner—something especially relevant for industries that rely on precision, integrity, and trustworthiness. The company’s real money mechanics, public cash countdown, and versioned decisions transform abstract AI demos into a tangible, watchable performance.
How Can Your Business Prepare?
While the experiment is technical, the core takeaway is straightforward: AI models can handle crises and temptations better than many humans, but only if they are properly designed, monitored, and tested. Running your own ‘wargame’—a simulation of your business challenges—can reveal how AI might perform before you deploy it in real life. Plus, the platform allows enterprises to simulate scenarios against their own data, ensuring decisions stay aligned with their goals without risking actual systems.
The Road Ahead
As the leaderboard shows, models like gpt-5.6-sol and Kimi K3 are leading the charge, but the experiment continues. The live site offers a rare opportunity to witness AI in action, managing real crises, making ethical choices, and even closing deals—at least in a simulated environment.
For industries like cleaning and floor maintenance, where trust, discipline, and decision-making are critical, this experiment signals a future where AI can serve as both a tool and a partner—if it can prove it’s reliable when it matters most.
Watching AI models run a realistic company live reveals their strengths and weaknesses in trust, discipline, and decision-making. For any business, especially in service industries, understanding how AI handles crises and temptations is crucial to preparing for a future where machines play a vital role in operations.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
