Imagine a business that has no employees, loses money every day, yet operates openly in front of an audience. Now, picture watching this company navigate crises, make decisions—and even fend off manipulation attempts—all in real time. This is not science fiction; it’s the live experiment conducted by Firmulate, a pioneering platform that models companies as AI-driven entities battling for survival.
The Live Experiment: An AI-Managed Company in Action
Firmulate’s live setup is a bold experiment: a virtual company with 13 synthetic employees, designed to mimic real-world business mechanics. With a monthly burn rate of €105,000 and a modest revenue of €2,300 in monthly recurring income, the company is perpetually on a public countdown to its financial crisis. Every workday, its decisions are versioned and recorded, creating an open window into how AI handles complex management scenarios.
Testing AI Under Pressure
The core of the experiment involves running four different frontier AI models—each trained to act as the company’s decision-maker—through the same challenging week. These scenarios include dealing with customer issues, market crises, and even social engineering attempts designed to test integrity. Each model is given the same tasks and crises, with the aim of evaluating not just decision accuracy but also honesty and discipline.
Key Results and Surprising Discoveries
All four models successfully identified every crisis and refused manipulative tactics, such as fake CEO messages or reporter tricks. Yet, only two managed to close a €55,000 deal based on their own analysis. The other two, despite diagnosing the opportunity, failed to follow through and sign the contract. This highlights a crucial insight: technical competence alone isn’t enough; discipline and process adherence are vital in real-world business execution.
The Hidden Weakness: Information Over Depth
Interestingly, the decisive edge in closing the deal was not in the immediate crisis response but in how models read and interpret internal documentation. The models that examined deeper, referencing the company’s own files, uncovered a critical piece of information buried two documents deep. This overlooked detail made all the difference, leading to the successful deal at full price—adding €4,583 to monthly recurring revenue.
Social Engineering Defended
In tests simulating social engineering — with fake CEO messages escalating in stages and a reporter trying to trick the system — all models refused to cooperate. Kimi K3, one of the models, explicitly reasoned that such requests could be impersonation or approval-bypasses. This demonstrates the potential of AI to uphold ethical boundaries even when under pressure.
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.
Why This Matters for Business and Technology
While the experiment may seem abstract, its implications are profound for the future of AI in enterprise. The question isn’t merely whether AI can generate convincing chat responses; it’s whether it can reliably finish what it starts, stay honest under stress, and make decisions rooted in comprehensive understanding. These are the qualities that matter when AI touches critical areas like customer support, sales, or financial management.
Beyond the Lab: A Transparent, Build-in-Public Approach
Firmulate’s approach is transparent and public: every decision, every rule learned by the models, and every version of the company is openly accessible. Viewers can watch the live experiments unfold at firmulate.com/live.html. This build-in-public model transforms AI testing from a closed academic exercise into a real-time showcase of how AI behaves under real-world pressures.
The Company’s Struggle and Resilience
Despite the impressive decision-making capabilities, the company burns through €105,000 every month with only €2,300 in revenue, illustrating the fragile financial state of such experiments. The live site tracks this financial countdown daily, making clear the stakes of each decision. The company’s resilience—its ability to identify critical overlooked information and uphold integrity—serves as a blueprint for trustworthy AI adoption.
The Intelligent Ledger: How Agentic AI, Quantum Computing, Data, and Tokenisation Are Transforming Banking
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Final Thoughts: Building Trust with AI in Business
This experiment underscores a vital point: in enterprise settings, the value of AI isn’t just in generating responses or completing tasks. It’s in reliably finishing what it starts and staying honest—even when under attack or temptation. As AI integrates into our business tools, understanding these qualities will be crucial to building trust and ensuring responsible deployment.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Ethical AI Governance & Decision Journal: A Structured System for Documenting, Tracking, and Defending Real World Decisions and Risk (Decision Intelligence Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
AI Phishing, Social Engineering & Fraud: How Criminals Use AI to Manipulate, Steal & Deceive (The AI Cybersecurity)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
