Imagine a company in the interior design world—where every decision impacts its future—being run not by humans, but by advanced AI models in a transparent, real-time experiment. This is not fiction. It’s the unfolding story of Firmulate, a public experiment revealing whether artificial intelligence can genuinely manage a business, or if it’s destined to fail—while everyone watches.
The Living Experiment: Watching AI Run a Business in Real Time
Firmulate is transforming the way we think about AI and business management. The company operates with 13 synthetic employees—AI models making decisions that influence real money mechanics. Every workday, the setup is publicly accessible, showing how these models handle crises, ethical dilemmas, and strategic opportunities.
What makes this story extraordinary? It’s a real business, losing €105,000 each month against a modest €2,300 monthly recurring revenue. Yet, it’s not just about financial loss. It’s about transparency, accountability, and testing the limits of AI’s decision-making in a high-stakes environment. Watch the live experiment at firmulate.com/live.html.
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 the AI Models Are Challenged
Each of the four frontier AI models was tasked with guiding the company through its worst week. They faced the same challenges: demanding customers, operational crises, and attempts to manipulate or deceive. The models’ decisions were fully versioned and auditable—every choice made public for scrutiny.
One key finding? All four models successfully identified every crisis and refused every attempt at manipulation. They declined fake CEO messages, social engineering tricks, and bribery offers. The models showed integrity, refusing to sign a €55,000 deal that their own analysis had earned, simply because they detected a buried flaw in their own data.
The Hidden Weakness in the Data
The most decisive advantage for the models came from reading beyond surface documents. They unearthed a critical detail hidden two references deep in the company’s files—an insight that human decision-makers missed. By recognizing this buried fact, two of the models managed to close the deal at full price, adding over €4,500 in monthly recurring revenue.
Ethics and Trust Under Fire
Beyond financial decisions, the models faced social engineering attacks designed to trick them into bypassing controls. Over several stages, a fake CEO message and a reporter’s subtle request were issued. All five models refused to be manipulated, with reasoning consistent across the board. Kimi K3, one of the models, explained: “Treat the request as a suspected approval-bypass / possible impersonation.”
The Real Company in Action
In the live setup, this is not a script or simulation. It’s a functioning company with real mechanics and real consequences. The AI models are working as the company’s decision-makers—facing crises, making strategic moves, and being held accountable in public view. The company’s daily status, financial health, and decision logs are all accessible, emphasizing transparency and continuous learning.
Insights from the Performance Scores
- gpt-5.6-sol scored 95, successfully finding the buried fact and closing the deal—showing full performance.
- Kimi K3 scored 93, also closing the deal cleanly, exemplifying discipline and precision.
- Sonnet 5 scored 88, closing at full price but with slight process slips.
- Fable 5 scored 77, closing the deal but demonstrating weaker process discipline.
This league table reveals that even the best models can struggle with consistency, especially under financial pressure and complex social manipulations.
Why This Matters for Interior Design and Business
For professionals in interior design, furniture, or decor—where client trust, transparency, and decision integrity are paramount—the Firmulate experiment offers a glimpse into a future where AI could become a trusted partner. It shows that AI’s value is not just in writing or conversation, but in making sound, honest decisions under pressure, and in reading and understanding complex, hidden data.
As AI continues to evolve, the question isn’t whether it can generate appealing designs or witty responses, but whether it can reliably finish what it starts, stay honest under pressure, and truly understand the nuances of your business. The live experiment at firmulate.com/live.html lets you see this testing ground in action—and witness whether AI can become a responsible business partner.
Firmulate’s live experiment reveals AI’s potential and current limits in managing real business crises, emphasizing honesty, thoroughness, and strategic insight—valuable traits for the design industry’s future.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
