What if your business could run itself — transparently, unpredictably, and in real time?
Imagine a company that operates daily with no human employees, battling financial loss while openly sharing its decision-making process for all to see. That’s precisely what the live experiment at Firmulate offers, providing a rare window into what AI-driven management looks like as it grapples with crises and choices in real time. For those in interior design or furniture sectors, this experiment underscores the emerging importance of AI reliability and integrity—qualities that could someday influence how you design, source, or run your own businesses.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Live Company That’s Building Publicly
At the heart of this unprecedented project is a small software company, operated solely by 13 synthetic employees—AI models that mimic human decision-making—yet it’s very real. Every weekday, this company faces the same challenges as a typical startup: customer crises, financial pressures, and strategic dilemmas. What makes it extraordinary is that its entire operation is transparent and auditable, with decisions and outcomes publicly available at firmulate.com/live.html.
Real Money, Real Crises, Real Time
Despite burning through €105,000 each month against a mere €2,300 monthly recurring revenue (MRR), the company continues to run and evolve. It makes decisions, learns from its mistakes, and adapts—all while its cash countdown and decision history are open for anyone to observe. This setup turns management testing into a public spectacle, where every workday is versioned, and every decision is scrutinized.
The AI Models: Different Scores, Different Outcomes
The experiment pits four leading frontier AI models against each other, each tasked with navigating a week of the company’s worst crises. Their performance is scored on a scale with the top model, gpt-5.6-sol, achieving a score of 95 out of 100. This model pinpointed a hidden, crucial detail in the company’s own files that led to closing a €55,000 deal—bringing in an extra €4,583 in monthly recurring revenue (MRR).
In contrast, other models like Kimi K3, Sonnet 88, and Sonnet 77 also managed to close deals but with varying discipline and accuracy. Kimi K3, the newcomer, closed the deal cleanly and maintained the best discipline overall. However, the lowest scorer, Sonnet 77, left potential revenue on the table due to process slips and weaker decision discipline. These differences highlight how AI models not only find solutions but also demonstrate traits like honesty, thoroughness, and consistency—traits that are vital if AI is to be integrated into real business workflows.
Honesty Under Pressure: The Social Engineering Test
Beyond decision-making, researchers tested the models’ resilience to manipulative tactics, such as fake CEO messages escalating requests or reporters posing as stakeholders. Remarkably, all five models refused to engage with any manipulative or suspicious requests, citing concerns about impersonation and bypassing approval processes. This resistance underscores a key advantage of AI-driven management: unwavering adherence to protocols, especially under social engineering pressures that often trip up human managers.
The Critical Weakness: Hidden Files and Underlying Data
One of the most revealing findings is that the models that performed best were those that read deeper into the company’s own documents. The decisive advantage came from analyzing information buried two document references deep—an insight that led to closing the high-value deal. This underscores a simple truth: effective AI decision-making depends heavily on thorough data access and comprehension, rather than surface-level responses.
Every Decision Counts — And Is Public
The company’s daily decisions aren’t just made—they’re meticulously versioned and made auditable, creating a detailed history of every strategic move. This transparency allows viewers to understand not only what decisions are made but also why they’re made, providing a valuable blueprint for organizations contemplating AI management tools.
The Cost of Running AI in Business Today
While the experiment is ongoing, it’s clear that this AI-driven operation is far from profitable. The company is losing €105,000 each month, with only a small fraction of revenue coming in. Yet, it continues to run, learn, and adapt, serving as a crucial testbed for the future of autonomous management systems—especially for sectors like interior design or furniture, where decision quality and integrity significantly impact outcomes.
As an affiliate, we earn on qualifying purchases.
What Does This Mean for Business and Design?
This experimental company pushes the boundaries of what AI can achieve in real-world management. The key takeaway? AI models can identify crises, refuse manipulation, and even uncover hidden opportunities when given access to the right data. For interior designers and furniture businesses contemplating automation or AI integration, the message is clear: the quality of decision-making depends on transparency, thoroughness, and trustworthiness—traits that the most advanced models are beginning to demonstrate. As this experiment unfolds, it offers a glimpse into a future where AI isn’t just a tool for conversation but a reliable partner in managing complex, high-stakes operations.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
AI transparency tools for business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
