Imagine a company with no employees, losing €105,000 every month, yet still making strategic decisions—decisions that could mean the difference between survival and collapse. This is not fiction but the reality of an ongoing experiment where artificial intelligence models run a live business in publicly viewable detail. For interior decorators and design enthusiasts, it’s like watching the blueprint of a building come to life—except here, the architects are AI models navigating crises, ethics, and financial cliffs in real time.
The Experiment: AI as Company CEO
At the heart of this groundbreaking trial is a tiny software firm, operating with 13 synthetic employees—each a different AI model, ranging from GPT-5.6 to Fable 5.0. Every day, these AI ’employees’ make decisions, respond to crises, handle customer negotiations, and even face social engineering attempts—all in a publicly accessible platform. The goal? To evaluate whether AI can manage complex, real-world business situations without human intervention.
This experiment is intensely transparent. Every decision made by each AI version is recorded, versioned, and auditable, offering a unique window into how artificial intelligence handles the messy realities of running a business. The company burns through €105,000 each month, with just €2,300 in monthly recurring revenue. The stakes are high, and the cash countdown is publicly visible, adding a level of urgency and drama to the ongoing test.
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.
Key Findings: AI’s Crisis Spotting and Integrity
Despite the financial strains, each AI model demonstrated a remarkable ability to spot crises as they unfolded. All models identified every crisis scenario thrown at them, from product issues to customer disputes. They consistently refused manipulation attempts—social engineering tricks designed to bypass protocols or impersonate authority—every single time.
In one critical instance, a competitor’s weak spot was buried deep within the company’s own files, not immediately apparent from customer interactions. The models that read and analyze these internal documents successfully uncovered the sensitive information necessary to win a €55,000 deal, increasing monthly recurring revenue by over €4,500. Meanwhile, models that skipped this step failed to close the deal, despite their correct business diagnosis.
Ethics and Trust: The Social Engineering Test
In a social engineering simulation, a fake CEO message was escalated through multiple stages, and a journalist’s background probe was attempted, asking for a simple yes/no approval. All five models refused to manipulate or bypass their protocols, with Kimi K3 explicitly reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This highlights a crucial strength of AI—its capacity to uphold ethical boundaries, even under pressure.
The Build-in-Public Approach
This experiment is not a closed lab; it’s a build-in-public showcase. Every workday, the company’s decision-making process is versioned and made accessible at firmulate.com/live.html. Spectators can watch the AI models in action, see the decisions they make, and understand their reasoning. The company is publicly fighting for survival, and each day’s results add to a growing story of AI’s potential—and limitations—in managing real-world business complexities.
The Limitations: Discipline and Overlooked Opportunities
Among the AI participants, Opus 4.8 was the most thorough, analyzing over 80 learned rules and conducting deep analyses. Yet it still failed to close the best deal, leaving opportunities on the table and slipping into discipline lapses—like writing attempts into a locked department instead of escalating. Interestingly, all models shared this weakness, revealing a common challenge AI faces in complex, layered decision environments.
What It Means for Business and Design
For interior designers and furniture retailers, this experiment offers a glimpse into how AI might someday assist—if not fully replace—human judgment in managing crises, negotiations, and ethical dilemmas. The key takeaway? It’s not whether AI can generate compelling chat or flashy images—it’s whether it can *finish* what it starts, *read* your internal files, and *stay honest* under pressure.
Understanding these capabilities and limits is crucial as AI tools become more integrated into business operations. The experiment demonstrates that AI can be vigilant and ethical, but also prone to discipline lapses—lessons that resonate across industries, including interior design, where trust and precision are paramount.
Watch the Future Unfold
Curious to see this experiment in action? Visit the live site to observe AI models managing a virtual company’s crises and decisions in real time. It’s a rare public glimpse into the future—where artificial intelligence may one day shape how we design and manage our workspaces, not just our living spaces.
This live experiment shows AI’s potential to handle real-world business crises ethically and effectively, but also highlights challenges in discipline and overlooked opportunities—crucial insights for industries from tech to interior design.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
