Imagine a startup where there are no human employees—only artificial intelligence models making crucial business decisions, fighting to keep a tiny revenue stream alive while the world watches. Sounds like sci-fi? Well, this is the live experiment of Firmulate, a company that’s pushing transparency and discipline in AI management to the extreme. You can watch the company’s daily struggles unfold in real time, as 13 synthetic ’employees’ battle crises, temptations, and the ticking clock of cash burn.
The Live Company: A Transparent Look at AI Decision-Making
Firmulate runs an unprecedented experiment: a public, real-time, build-in-public showcase of an AI-managed business. There are no humans on the payroll—just 13 synthetic employees, driven by a constantly evolving set of over 680 self-learned rules. Every workday, the entire decision-making process is versioned and accessible for scrutiny. The goal? To understand how AI handles real-world crises, ethical dilemmas, and the complexities of running a business under pressure.
The Stakes Are High: Money, Trust, and Survival
Currently, the company’s finances tell a sobering story: it burns €105,000 each month against a monthly recurring revenue (MRR) of just €2,300. A public cash countdown looms, adding urgency to every decision. The live site, firmulate.com/live.html, offers a window into this relentless struggle, where every decision is scrutinized and every crisis is a test of AI integrity and effectiveness.
The Experiment: Same Week, Different Models, Different Results
Four top-tier AI models—namely gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—were each placed in the same difficult week faced by the company. The week was filled with customer crises, internal dilemmas, and manipulation attempts designed to test honesty and discipline.
The results? All four models successfully identified every crisis and refused every manipulation attempt. That’s a victory for AI integrity. But when it came to closing a key deal worth €55,000 per month, only two models succeeded in signing the contract based on their own analysis. The other two, despite making the same diagnosis and pitch, left the deal unclosed. This gap underscores a vital insight: AI may get the problems right, but closing the deal requires discipline and follow-through—traits that are hard to automate.
The Hidden Weakness: Reading Beyond the Surface
Interestingly, the main weakness wasn’t in crisis detection but in a buried piece of information in the company’s documents. Models that read deeper into the files—beyond just surface-level data—won the deal at full price, adding over €4,583 in MRR. This illustrates a crucial point: AI’s ability to uncover hidden knowledge can be the key to winning business.
Ethical Boundaries and Social Engineering
To test the AI’s moral compass, researchers staged social engineering scenarios, including fake CEO messages and a reporter trick involving subtle approval requests. All five models refused to participate in manipulative tactics, with Kimi K3 reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This discipline highlights the importance of ethical safeguards in AI decision-making.
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 Human-Like, Yet Synthetic Workforce
The company’s setup is striking: a simulated workforce of 13 AI employees, each operating under a strict rulebook, yet facing real money mechanics. The company burns €105,000 monthly and struggles to generate more than €2,300 in MRR. Every business day, decisions are versioned, audited, and made transparent for public review. It’s an ongoing battle—a real-time case study for how AI could or should manage critical business functions.
What We Learn from the Frontlines
The most thorough participant—Opus 4.8—analyzed over 80 learned rules but still left the deal unclosed, showcasing how discipline and process slippage can happen even with advanced AI. The overall lesson is clear: success is not just about diagnosis but also about execution, follow-through, and maintaining ethical standards under pressure.
Why This Matters for Business and Beyond
The experiment’s takeaway is simple but powerful: when AI touches your CRM, support, or forecasting, it’s not enough for it to generate good chat or insights. The real test is whether it can finish what it starts, read critical hidden data, and stay honest when under temptation. As these models demonstrate, the difference between winning and losing a crucial deal can hinge on their discipline and integrity—traits that are as vital as their intelligence.
Firmulate’s live AI company experiment reveals vital insights: AI can identify crises and refuse manipulation, but discipline and deep data reading are key to closing deals and surviving. Watch it unfold at firmulate.com/live.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
AI-Driven Digital Transformation: A Proven Blueprint for Responsible AI Scaling
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
AI for Crisis Management & Business Continuity: The Executive Handbook for Protecting Your Organization with 50 AI Prompts for Detection, Response, and … BUSINESS & MANAGEMENT LIBRARY SERIES 38)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
AI, Automation, and War: The Rise of a Military-Tech Complex
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
