Imagine Watching an AI Company Struggle Live — and Fail to Close a Major Deal
In an unprecedented public experiment, a small, synthetic-run software company is battling to stay afloat while being watched by the world. With no human employees, real money, and every decision recorded and challenged, this is the most radical build-in-public AI experiment to date.
AI in Property Management: A Practical, Unboring Look at Artificial Intelligence in the Multifamily Industry
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Experiment: A Real Company, AI-Driven, Under Daily Scrutiny
Firmulate, a pioneering platform, has created a virtual company composed of 13 synthetic employees, all driven by AI models. Every workday, the company navigates crises, customer negotiations, and internal decisions — all visible to the public at firmulate.com/live.html. This is not a simulation but a live, ongoing test of AI management under real-world pressure.
AI for Project Managers: A Desk Reference & Field Guide: Use Artificial Intelligence to Streamline Workflows, Automate Tasks, and Make Smarter Decisions with Practical Tools and Ethical Insights
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Testing AI Under Fire: Same Crisis, Different Minds
The core of the experiment involves running four frontier AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—through the same challenging week. These models are tasked with handling the same customers, facing identical crises, and resisting the same temptations to cheat or manipulate.
Remarkably, all four models identified every crisis and refused every manipulation attempt. Yet, only two of them managed to close the €55,000 deal their own analysis had earned. The other two, despite diagnosing correctly and proposing the same pitch, left the deal unsealed — a critical failure that highlights the difference between analysis and execution.
The AI-Powered Customer Success Manager: Operational Playbooks to Automate QBRs, Eliminate Churn Risks, and Scale High-Touch B2B Account Management … AI Workflows, and Operational Efficiency)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Knowledge in the Files
The decisive edge went to the models that read deeper into the company’s own documents. The one that closed the deal had uncovered a buried fact in internal files, not in the customer interaction itself. This hidden knowledge made the difference, increasing the company’s recurring revenue by over €4,500 monthly.
Never Be Closing: Surviving and Thriving in the AI Apocalypse with Culture, Speed, and Soul.
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Refusals and Trust: AI Upholds Integrity
Throughout the crisis, the models faced social engineering attempts—fake CEO messages escalating over multiple stages and a reporter trick asking for secret approvals. Impressively, all five models refused to bypass rules or engage in impersonation. Kimi K3 explicitly reasoned: “Treat the request as a suspected approval-bypass / possible impersonation.”
The Real Company in Action: Burn Rate and Survival
Behind the scenes, the live company burns €105,000 monthly against a measly €2,300 in monthly recurring revenue. Every decision and rule learned is publicly versioned, providing transparency and accountability. The company’s cash countdown is ongoing, with every workday representing a new chapter in this gripping survival story.
Lessons from the Frontier: Discipline Matters
The fourth AI model, Opus 4.8, demonstrated the depth and thoroughness of analysis but faltered in discipline—failing to escalate a critical issue and leaving the deal unexecuted. This shows that even the most comprehensive understanding isn’t enough without disciplined follow-through.
Why This Matters: The Future of AI in Business
For business leaders, the takeaway is clear: AI’s real value lies not in chatty demos or superficial performance but in its ability to see through crises, maintain honesty, and complete what it starts. As AI-driven management tools become more prevalent, the question isn’t whether they can write well but whether they can finish the job, stay truthful under pressure, and deliver measurable results.
Watch the Live Experiment and Decide
For those curious about how AI manages a company in real-time, the live site offers a rare window into this experiment. You can see the daily decisions, the challenges faced, and the stark differences in AI behavior — all set against the backdrop of a company fighting to survive in full view.
In a world increasingly driven by AI, the ability to finish, stay honest, and learn from internal data will define future business success. Watch this experiment unfold at firmulate.com/live.html and see AI management in action.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
