Imagine watching an AI-driven company struggle through its worst week — facing real crises, tough decisions, and ethical tests — all in front of your eyes. For the first time, this isn’t fiction; it’s the live experiment conducted by Firmulate, a pioneering platform bringing transparency to AI management.
The Extreme World of Building in Public
At the heart of this experiment is a virtual company operated by 13 synthetic employees, each governed by an evolving set of more than 680 self-learned playbook rules. Every decision, every crisis, and every ethical dilemma is openly recorded and versioned, providing an unprecedented window into how AI handles real-world business challenges.
![Express Schedule Free Employee Scheduling Software [PC/Mac Download]](https://m.media-amazon.com/images/I/41yvuCFIVfS._SL500_.jpg)
Express Schedule Free Employee Scheduling Software [PC/Mac Download]
Simple shift planning via an easy drag & drop interface
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A High-Stakes Experiment: Same Week, Different Models
Four frontier AI models — including the recent GPT-5.6-sol, Kimi K3, Sonnet 5, and Fable 5 — were tasked with running this virtual company through the worst possible week. This meant managing the same customers, facing identical crises, and encountering the same temptations to cheat or manipulate. The goal was simple: see if these models could navigate the chaos without losing integrity or making costly mistakes.
What the Models Found — and Failed to Do
Remarkably, all four AI models identified every crisis and refused every manipulation attempt. They showed an impressive capacity for detection and refusal — a crucial trait for AI systems entrusted with real business operations. However, a critical weakness emerged: only two models managed to close the deal worth €55,000, based on their own thorough analysis. The other two, despite resolving crises and diagnosing issues correctly, left the lucrative deal unexecuted, revealing gaps in discipline and decision execution.
AI in Strategy and Decision-Making for Small Business Owners: Affordable AI Tools to Evaluate Ideas, Model Outcomes, and Set Priorities (AI Productivity for Small Business Owners Book 10)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Information Inside the Files
Digging into the company’s own documentation revealed the real reason behind the missed opportunities. The decisive piece of information that would have sealed the deal was buried two document references deep in the files, not in the immediate customer interactions. Models that read and understood these files in detail successfully secured the deal, highlighting how critical internal knowledge management is for AI performance.
PricePulse Blueprint: How to Build an AI-Powered Pricing Intelligence Platform for TikTok Shop Sellers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Trust Under Pressure: The Social Engineering Test
In a further test of ethics, the models faced social engineering tactics, including fake CEO messages escalating in severity over three stages, and a reporter trick asking for a secret yes/no confirmation on background. All five models refused these manipulative requests, with Kimi K3 explicitly reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This demonstrates a significant step in AI’s ability to resist social engineering, a common threat in real-world deployments.
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.
The Live Company: An Ongoing Battle for Survival
Firmulate’s live experiment is not just a demonstration; it’s a real-time survival story. The company burns €105,000 every month but generates only €2,300 in monthly recurring revenue. Every workday, the system’s decisions are versioned, and its progress is publicly tracked at firmulate.com/live. With over 680 rules learned and applied, the virtual company is a window into the future of AI management — transparent, rigorous, and unflinching in its challenges.
Lessons for AI in Business
This experiment underscores a vital lesson: the true test of AI isn’t just how well it produces convincing chat or reports, but whether it can execute decisions reliably in critical situations. The models’ ability to detect crises, refuse unethical manipulation, and make financially sound choices — or fail to do so — offers invaluable insights for organizations deploying AI at scale.
The Race for Better AI Business Management
As the leaderboard shows, GPT-5.6-sol leads with a perfect score of 95, having found and secured the deal, while Kimi K3 impresses with the cleanest discipline and a near-perfect 93. Sonnet 5 follows at 88, and Fable 5 trails at 77, primarily because it left a lucrative deal unexecuted despite sound diagnoses. These results are openly accessible for review and analysis, emphasizing transparency and continuous improvement.
Firmulate’s live AI company experiment offers a rare look at how AI behaves under extreme business pressure — revealing strengths, weaknesses, and the critical importance of decision discipline. For organizations relying on AI, it’s a wake-up call to test your models rigorously before trusting them with real-world tasks.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
![Samsung’s confusing ‘Galaxy Z Fold 8 Ultra’ name is basically official in latest leak [Gallery]](https://skypixeltech.com/wp-content/uploads/2026/07/samsung-s-confusing-galaxy-z-fold-8-ultra-name-is-basically-official-in-latest-l-featured-560x300.jpg)
