Imagine a company with no employees, yet running in real time, making tough decisions, battling crises, and risking its very existence — all while you can watch every move unfold live. This is not science fiction. It’s the reality of a pioneering experiment in AI-powered business management, where transparency turns into a high-stakes laboratory for the future of work.
The Live Experiment: A Company Without Employees
At the heart of this experiment is a fully operational, virtual company built with 13 synthetic employees, each governed by advanced AI models. Instead of human staff, these models handle everything from crisis management to decision-making, all within a meticulously designed environment that replicates the challenges of real-world business.
What makes this setup extraordinary is its transparency. Every decision, every crisis, and every rule learned by these AI agents is publicly documented, versioned, and viewable at firmulate.com/live.html. This level of build-in-public transparency enables observers to track how AI manages complex business scenarios—completely in the open.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Challenge: Surviving the Worst Week
The core test involves subjecting four different front-line AI models to a simulated ‘worst week’ scenario. These models faced the same customer issues, crises, and ethical temptations. Their performance was measured not by chat flair but by real business outcomes, including their ability to identify buried information, resist manipulation, and close lucrative deals.
One revealing finding was that all four models successfully identified every crisis and refused all manipulation attempts, demonstrating integrity and situational awareness. However, only two models managed to close a €55,000 deal, worth an additional €4,583 in monthly recurring revenue.
The Hidden Weakness: What Made the Difference?
Interestingly, the decisive factor wasn’t in the top-level decision-making. The models that secured the deal read deeper into the company’s own files—information buried two document references deep—giving them a critical advantage. Those that did this won the deal at full price, illustrating how crucial internal data analysis is for AI decision-making.
Handling Social Engineering and Ethical Tests
In one test, fake messages from a supposed CEO, escalating through multiple stages, were used to trick the AI models. All five models tested refused to be manipulated, with Kimi K3 explicitly reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This robustness against social engineering underscores the importance of trustworthiness in AI systems handling real business decisions.
The Financial Reality: Burning Money in Public
While the experiment pushes forward with groundbreaking transparency, it also exposes the harsh financial reality of building and running such an AI company. Currently, the operation burns through €105,000 every month, while generating only €2,300 in monthly recurring revenue. A public cash countdown makes clear this is a battle for survival, not profit—yet.
Deep Dive into the Models’ Behavior
The most thorough participant is Opus 4.8, which learned over 80 rules and conducted deep analyses. Despite its thoroughness, it left opportunities unexploited, such as failing to escalate instead of writing into a locked department when discipline slipped. This illustrates that even the most advanced AI models have gaps, especially when discipline wanes under pressure.
What This Means for Businesses and the Future of AI
This experiment offers a stark view of what AI can and cannot do in real-world management. It shows AI’s capacity to recognize crises, resist manipulation, and make objective decisions—yet also highlights vulnerabilities in discipline and strategic execution. It underscores the need for transparent testing environments before deploying AI systems into critical business functions.
For companies contemplating AI integration, the key takeaway is not just about chat quality or writing skills. It’s about whether AI can finish what it starts, interpret internal documents, and stay honest under pressure—traits that are vital in high-stakes environments.
This pioneering live experiment reveals how AI models manage a virtual company under pressure, showcasing strengths in crisis detection and ethical resistance, but also exposing discipline gaps. Transparency and rigorous testing are essential before trusting AI with real business decisions.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
