What if your smart home devices faced the harsh realities of running a business — every day, under real pressure?
Imagine an AI-powered company that operates without employees, losing money every month, yet still navigating crises, making decisions, and even closing deals — all while being fully transparent and public. This isn’t science fiction; it’s the live experiment conducted by Firmulate, showcasing how AI models handle the complexities of managing a real business in real time.
As an affiliate, we earn on qualifying purchases.
The Firmulate Live Company: An Unprecedented Business Experiment
At the heart of this experiment is a small, software company managed entirely by AI models. Unlike typical chatbots or virtual assistants, these models are tasked with running a company through its worst week — facing customer crises, ethical dilemmas, and strategic decisions. Every move they make is tracked, versioned, and open to public scrutiny at firmulate.com/live.html.
This company has 13 synthetic employees, each following a complex set of 680+ self-learned rules. Despite burning €105,000 monthly with only €2,300 in monthly recurring revenue, it continues to operate, with a public countdown to insolvency. The purpose isn’t just to test AI’s decision-making but to see how well these models can replicate genuine management challenges.
As an affiliate, we earn on qualifying purchases.
What Do the Results Reveal About AI’s Business Capabilities?
Four leading AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8—were each tested on the same business week, with identical crises, customer demands, and temptations. Remarkably, all four models recognized every crisis and refused every manipulation attempt, such as social engineering tricks or fake CEO messages. However, only two successfully closed the deal worth €55,000, earning full revenue. The other two failed to follow through, despite making the right diagnosis and pitching the deal.
Digging deeper, the real weakness was hidden in the company’s own files, not in customer interactions. The models that reviewed and understood the deeper documentation managed to win the deal at full price, highlighting the importance of thorough information reading and comprehension.
This experiment underscores a crucial point for businesses: AI’s ability to identify crises and uphold honesty is impressive, but execution — actually closing deals and sticking to discipline — remains challenging. The company’s real struggle is not in spotting problems but in resolving them consistently, especially under pressure.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
