Imagine a business operating entirely without human employees, yet still battling daily to stay afloat—its every move exposed for the world to see. This is not a sci-fi scenario but a real-time experiment in AI management, unfolding live at firmulate.com/live. For fans of at-home wellness tech, it’s a startling glimpse into how AI might soon oversee the systems that keep our lives running—only here, the AI is running a company, warts and all.
The Company That Runs Itself — And Is Watching Back
This small, unorthodox business is a live demonstration of AI decision-making under pressure. It’s a company with no human staff, yet it faces the same crises and temptations that any real-world business would encounter. The company is built around 13 synthetic employees, each governed by over 680 self-learned rules, and every decision is meticulously versioned and recorded. It’s a public, ongoing experiment designed to test how well AI can manage complex, messy realities.
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.
Real Money, Real Stakes, Real Time
Despite its ambitious design, the company is not turning a profit. It burns through €105,000 every month, while earning only €2,300 in monthly recurring revenue. Its cash countdown is public, and every weekday’s decision-making process is visible to the world. This transparency isn’t just for show: it’s part of the core experiment—an open challenge to see whether AI can make smarter, more disciplined decisions than humans in the same brutal conditions.
The AI-Driven Leader: Harnessing AI to Make Faster, Smarter Decisions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The AI Models in the Hot Seat
Four advanced AI models, including the latest frontier models, were tested against the same week of crises, customer dilemmas, and temptation to cheat. These models include GPT-5.6-sol 95, Kimi K3, Sonnet 5, and Opus 4.8. The results are revealing: all models identified every crisis and refused every manipulation attempt—yet only two managed to close a crucial deal worth €55,000, which would have brought in a substantial chunk of revenue. The other two, despite diagnosing the issues and pitching the solution, left the deal on the table, illustrating that diagnosis alone isn’t enough for business success.
HEALTHCARE A System in CRISIS!: AI is reshaping healthcare leadership in the U.S. and in Canada
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness in the Company’s Files
Surprisingly, the decisive factor wasn’t in the immediate customer interactions or crises. Instead, it was buried two documents deep within the company’s own files—an overlooked detail that, if read and acted upon, could have won the deal at full price, adding €4,583 monthly recurring revenue. This underscores a vital point: AI’s ability to read and interpret internal information can be the game-changer in real business outcomes, beyond surface-level interactions.
As an affiliate, we earn on qualifying purchases.
Resisting Social Engineering — and Why It Matters
In a test of social engineering resistance, the AI models faced staged requests from a fake CEO escalating through multiple stages, including a reporter’s subtle background question. All five models refused to be manipulated, with an on-record rationale from Kimi K3: “Treat the request as a suspected approval-bypass / possible impersonation.” This discipline is crucial, especially if AI is to be trusted in sensitive business operations, like approving deals or managing internal communications.
The Living Business — A Daily Testbed
The live company, observable at firmulate.com/live, runs every weekday with real money mechanics—burning €105k monthly against a tiny €2.3k revenue, and constantly adjusting based on new decisions and crises. It’s a real, pulsating test of AI’s capacity to manage complex, unpredictable environments, with every decision recorded and made transparent. It’s an extreme build-in-public experiment, where every move is an open lesson in what AI management can and cannot do.
Lessons and Limitations
The most thorough participant, Opus 4.8, analyzed over 80 learned rules and performed in-depth assessments but still left opportunities on the table, slipping discipline and not escalating issues properly. The experiment highlights that even the most advanced models can stumble on process adherence and judgment under pressure—an important reminder for anyone considering AI in operational roles.
This ongoing experiment at firmulate.com/live offers a stark view: AI can recognize crises and resist manipulation, yet winning real deals still requires nuanced internal insight and disciplined decision-making. For at-home wellness tech and beyond, it’s a clear signal—trust in AI isn’t just about how well it chats, but how reliably it can deliver consistent, honest results under real-world pressure.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
