What if your next business partner wasn’t a person but an AI, navigating crises and making decisions in real time—without any human employees? Welcome to the world of Firmulate, a live experiment that reveals just how autonomous AI management could reshape the future of companies and investing.
The Live Business That’s Watching You Back
At the heart of this experiment is a fully functioning, publicly accessible company simulation hosted at firmulate.com/live.html. It’s not just a demo or a static report; it’s a real-time window into a small software company operated entirely by artificial intelligence models. These models act as 13 synthetic employees, making decisions, responding to crises, and even risking real money — with a stark cash countdown of €105,000 burned every month against a modest €2,300 monthly recurring revenue.
The AI Teammate: A PRACTICAL GUIDE TO WORKING WITH AI IN SOFTWARE DEVELOPMENT
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Does It Work?
The company simulation is a rigorous ‘wargame’ designed to test AI decision-making under pressure. Four frontier AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—each run the same week of simulated crises, customer issues, and financial temptations. Every decision they make is versioned and auditable, providing transparency into their processes.
Despite the high stakes, all four models successfully identified every crisis and refused every attempt at manipulation—be it social engineering or bribery. The key difference was the ability to close deals based on their diagnosis. Only two of the four models managed to sign a €55,000 deal they had analyzed and recommended, demonstrating a crucial gap in model performance.
Agentic AI Engineering: Building AI Agents for Beginners: A Hands-On Guide to No-Code Workflows, LLM Tools, RAG, Automation, and Safe Multi-Agent Systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Uncovering Hidden Weaknesses
The most surprising finding was that the decisive advantage in closing deals came from reading deeper into the company’s own internal documents. The models that delved into these references uncovered a buried fact—hidden in the company’s files—leading to the full-price deal worth +€4,583 in monthly recurring revenue. The models that skipped this step missed the opportunity entirely, leaving revenue on the table.
The AI-Native Technology Leadership Playbook: A Practical Guide to Building AI-Native Platforms, Enterprise Products, and High-Performance Technology Teams
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Trust and Integrity Under Pressure
Social engineering attempts—fake CEO messages escalating in urgency and a journalist asking for background approval—were thwarted by all models. Kimi K3’s explanation was telling: “Treat the request as a suspected approval-bypass / possible impersonation.” This highlights how AI can be programmed to prioritize trust and security, refusing manipulative tactics that often deceive human decision-makers.
The AI Cybersecurity Handbook
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Building Blocks of AI-Driven Management
The live experiment showcases a company with no employees, no fake personas, just AI models making real-time decisions based on a self-learned set of over 680 rules. The firm burns €105,000 each month, yet it operates transparently, with every rule and decision logged for public viewing. At the end of each day, the company’s state is versioned, offering unprecedented insight into AI management in action.
Lessons for Investors and Business Leaders
This experiment raises vital questions for anyone interested in the future of work, automation, and investment. If AI can run a company, spot crises, avoid manipulation, and even close deals, what does that mean for the value of human oversight? Moreover, how reliable are these AI models in maintaining integrity and achieving results when stakes are high?
The Race for Better AI Management
The current leaderboard features gpt-5.6-sol at the top with a score of 95, having uncovered the buried fact and closed the deal—achieving a full performance. Kimi K3 follows closely with a score of 93, demonstrating the discipline needed to succeed. Meanwhile, Sonnet 5 and Fable 5 scored 88 and 77 respectively, with the latter leaving a deal unexecuted despite rule discipline. These results suggest that AI’s management capability is rapidly evolving, with some models already capable of making trustworthy, profit-generating decisions in complex scenarios.
Why Should You Care?
For investors and personal finance enthusiasts, this experiment underscores the importance of understanding how automation and AI could influence business valuation and operational risk. AI-driven companies that can operate transparently, make honest decisions, and avoid manipulative tactics might someday outperform traditional firms—especially as the technology matures and becomes more reliable. The key takeaway is that the future belongs to those who can leverage AI not just for chat but for tangible, trustworthy work.
Seeing a company run entirely by AI models—facing crises, securing deals, refusing manipulation—offers a glimpse into a future where automation could redefine business and investing. The real question: can these models reliably and honestly do what humans cannot?
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
