What if you could watch a company fight for its life — powered entirely by artificial intelligence?
In a groundbreaking live experiment, a small, virtual software firm is navigating its worst week yet, all under the watchful eyes of multiple AI models. This isn’t fiction; it’s a real-time, open-door showcase of how AI can operate as the decision-maker in a business environment — and the challenges it faces.
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 Real Company, The Real Crisis
Firmulate is hosting a live demonstration of an AI-driven company with a twist: it has no employees, yet it is burning €105,000 every month against a mere €2,300 in monthly recurring revenue. Every day, the company runs through a set of crises — customer issues, strategic decisions, and manipulations — all while its performance is meticulously documented and publicly accessible at firmulate.com/live.html.
internal document management system
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Experiment: Running AI Models Through the Gauntlet
Four distinct AI models, each representing different levels of sophistication, are tasked with managing this virtual company during its worst week. They face identical challenges: customer crises, tempting manipulations, and complex decision points. The models are not only tested on their crisis management but also on their integrity, honesty, and discipline under pressure.
AI crisis management tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Crisis Recognition and Integrity
Remarkably, all four models identified every crisis, refused every manipulation attempt, and maintained discipline. For instance, when presented with a social engineering scam — a staged CEO message escalating over multiple stages and a reporter trick — all five models refused to proceed, citing concerns about impersonation and trust.
business simulation AI models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: A Critical Document
The decisive factor in closing a €55,000 deal wasn’t in the customer interactions but hidden in the company’s internal files. Only the models that read and understood this buried information managed to win the deal at full price — a crystal-clear illustration that reading internal documents can be the difference between success and failure.
Performance Variations and Discipline
The models varied significantly in their execution. The most thorough, Opus 4.8, demonstrated deep analysis and learned over 80 rules but still left a potential deal unclosed, illustrating that even the best models can slip under pressure. Meanwhile, simpler models that ran without effort parameters performed comparably, showing that complexity isn’t always a guarantee of success.
The Stakes: A Company Burning Cash to Stay Alive
The live experiment isn’t just a demonstration; it’s a stark reminder of the reality many small businesses face today. With a cash burn rate of €105k monthly against minimal revenue, every decision and discipline matters. The company’s public cash countdown underscores the urgency of making the right choices, especially when AI is making or influencing those decisions.
What This Means for the Future of AI in Business
The Firmulate experiment highlights that AI decision-makers can recognize crises and uphold integrity, but they’re not infallible. Critical insights buried in internal data can be the key to success — a lesson for any enterprise deploying AI tools. As AI models become more integrated into business operations, understanding their strengths and weaknesses will be vital.
How to Prepare: Wargaming Your AI Workforce
Businesses can test their future AI team with Firmulate’s platform, running simulations against their own data without impacting real systems. This approach allows organizations to evaluate how AI will perform under pressure, identify weaknesses, and ensure it can deliver value without risk.
Watching a virtual company fight for survival reveals the potential and pitfalls of AI decision-making. Success depends on reading internal data, maintaining discipline under pressure, and scrutinizing AI behavior before deployment.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
