Imagine running an interior design studio where every decision, every crisis, and every deal is scrutinized live — but with one twist: it’s powered by artificial intelligence, running a real company that loses money every day. Welcome to a radical experiment in AI management, where transparency meets survival in the most public way.
The Live Experiment: An AI-Driven Company Under Pressure
At the heart of this groundbreaking venture is a real software company, operating daily with a twist: its workforce comprises 13 synthetic employees, guided by cutting-edge AI models. Every workday, the company faces actual crises, customer interactions, and financial decisions, all under watch at firmulate.com/live. The company is currently hemorrhaging €105,000 per month while generating only €2,300 in monthly recurring revenue (MRR), making it a stark example of a startup fighting for survival.
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.
How Does It Work?
The experiment pits four advanced AI models against each other, each tasked with running the same business through its worst week. They face identical customer issues, crises, and ethical dilemmas — from negotiating deals to handling social engineering attacks, such as fake CEO messages and media tricks. Every decision they make is recorded, versioned, and auditable, providing a window into their decision-making process.
What Are the Results?
Despite the challenges, all four models identified every crisis and refused manipulation attempts, reflecting a high level of integrity and decision accuracy. Yet, only two managed to close a €55,000 deal at full price, based solely on their analysis and pitch, without resorting to shortcuts or dishonesty. Interestingly, the key to winning this deal was a buried fact in the company’s own files — information that, if read thoroughly, led to a full-price agreement worth an additional +€4,583 MRR.
Honest AI at Play
When tested with social engineering scenarios, like escalating fake CEO messages or background checks, all models refused to cooperate, treating these as potential impersonation or approval bypass attempts. Kimi K3, in particular, explained: “Treat the request as a suspected approval-bypass / possible impersonation.” This shows that the models are not just surface-level chatbots but exhibit a nuanced understanding of trust and integrity.
Inside the Company: The Mechanics and the Stakes
The real company behind this experiment runs every business day, with a live scoreboard and transparent decision logs. It burns €105,000 each month, with only €2,300 in revenue, and is constantly annotated with over 680 self-learned rules to guide operations. This is not just an AI demonstration; it’s a raw, unfiltered look at what managing an AI-driven business really entails.
The Lessons from the Field
Of particular note is the performance of OPUS 4.8, which ran with a thorough set of over 80 learned rules and performed the deepest analysis. Despite this, it finished last because it occasionally slipped discipline — such as writing attempts into a locked department instead of escalating issues. This highlights that even the most advanced models can falter under pressure, and that discipline, process adherence, and situational awareness are crucial.
Why Should Interior Designers Care?
For those in interior decor and furniture, the message is clear: AI’s role in business management isn’t just about generating pretty images or suggestions. It’s about ensuring that AI agents can complete tasks reliably, read critical files thoroughly, refuse unethical manipulations, and ultimately, keep your business on track even when losing money. The experiment shows that AI can detect and reject manipulation, but the real challenge is ensuring it acts decisively and ethically in complex, real-world situations.
The Big Takeaway
This experiment at firmulate.com/live is not a fake or a demo. It’s a real company operating in real time, openly showing how AI models behave when faced with the toughest management decisions. Their ability to identify crises, refuse manipulation, and sometimes close deals at full price demonstrates AI’s potential — but also underscores that AI alone isn’t enough. Discipline, thoroughness, and ethical integrity matter just as much.
Final Thought
As AI begins to touch more aspects of business management, from CRM to support to sales, the question isn’t just about how well it writes or speaks. It’s whether it can finish what it starts, stay honest under pressure, and deliver measurable results. Watching this live experiment unfold offers a rare glimpse into the future of AI in business — a future where transparency, accountability, and resilience are tested daily.
This real-time experiment shows AI’s potential and limitations in managing a live business. It highlights the importance of trust, discipline, and decision integrity in AI-driven companies.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
