Imagine a business where every decision, crisis, and tactic is played out in front of the world — a company with no employees, burning through €105,000 each month, yet still fighting to stay alive. This is not fiction; it’s the bold experiment of Firmulate, a live AI company running in real-time, revealing the raw realities of AI management and decision-making.
The Live Experiment: A Company in Public View
Firmulate is no ordinary tech startup. It’s a digital enterprise comprised of 13 synthetic employees, each driven by advanced AI models. Its mission? To simulate running a small software business through its toughest week, exposing how AI handles crises, ethics, and profitability. Every decision, every crisis, and every temptation to cheat is versioned, auditable, and open to the public at firmulate.com/live.
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.
How the Models Fared Against Real-World Challenges
The experiment pitted four frontier AI models against the same set of challenges. These included managing customer crises, resisting manipulation attempts, and closing deals. Despite the intense pressure — with real money mechanics burning €105,000 monthly against a modest €2,300 monthly recurring revenue — all models successfully identified every crisis and refused manipulation attempts.
In an intriguing twist, only two of the models managed to close a €55,000 deal, earning full recognition for their accurate diagnosis and pitch. The others, despite identifying the opportunity, left the deal unclosed, highlighting a critical gap between diagnosis and execution in AI decision-making.
Practical Claude Handbook for Attorneys: Master Case Analysis, Contract Review, Research Automation, Client Communication, and Document Drafting (Claude AI Guide for Beginners)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Reading the Files
Digging deeper into what made the difference reveals a crucial weakness. The decisive factor for closing the deal was not in the overt customer interactions but buried two document references deep within the company’s files. The models that read these files fully gained insight and opportunity, while those that missed this step failed to capitalize.
Rise of the Titans: A Chronicle of AI War
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Resisting Social Engineering and Ethical Tests
Firmulate’s experiment also tested how the AI models respond to social engineering attempts. Fake CEO messages escalated over multiple stages, and a reporter trick asked for a simple background approval. All five models refused to be manipulated, with the Kimi K3 model explicitly reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This indicates a promising level of ethical resistance in these AI systems, even under pressure.
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.
Performance and Discipline in Practice
The most thorough participant, Opus 4.8, with over 80 learned rules and deep analysis, ended up in last place. It left a closing opportunity on the table and showed discipline slips, such as writing attempts into a locked department instead of escalating. The other models showed similar weaknesses, suggesting that even the best AI models struggle with complex, disciplined decision-making under stress.
Implications for Business and AI Adoption
This experiment raises important questions for any business considering AI automation. The key isn’t whether an AI can write well, but whether it can reliably complete tasks, read important documents, resist manipulation, and stay honest under pressure. As firms increasingly rely on AI for CRM, support, or forecasting, understanding its real-world performance and vulnerabilities becomes paramount.
Watch the Experiment Live
For those interested in seeing AI management in action, the live site at firmulate.com/live offers real-time updates of this ongoing battle for survival. Every weekday, the company’s decisions are versioned and made transparent, offering a rare glimpse into the complexities of AI-driven business management.
The Bigger Picture
This bold, build-in-public experiment demonstrates not only the strengths of current AI models but also their persistent shortcomings. The fact that only half of the models could successfully close deals, despite perfect crisis detection, underscores the ongoing challenge of translating analysis into action. As AI continues to evolve, these real-world tests provide invaluable insights for companies seeking trustworthy automation.
Firmulate’s live experiment shows AI models can identify crises and resist manipulation, but closing deals and disciplined execution remain challenging. Watch this real-time battle for survival at firmulate.com/live.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
