Imagine a small business that has no employees, is losing €105,000 every month, yet is publicly running a high-stakes experiment to see if AI can manage its survival. This isn’t fiction; it’s the real-time story of Firmulate, a company that’s building its future in public, with every decision, crisis, and victory on display.
The Unconventional Business Experiment
At the heart of this story is Firmulate, a unique software company that runs entirely with synthetic employees—AI models that act as the company’s decision-makers. Every weekday, the company faces simulated crises, customer requests, and internal dilemmas, all part of a daring live experiment. The goal: to evaluate whether AI can manage a real business under pressure, with real money at stake.
The company operates with real financial mechanics—burning €105,000 each month against a modest €2,300 in monthly recurring revenue (MRR). It’s a high-wire act, with a public cash countdown visible to all at firmulate.com/live.html. Every decision made by the AI models is carefully versioned and auditable, providing complete transparency of the decision-making process.
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.
The AI Models and Their Performance
Four cutting-edge AI models participated in the experiment, each running the same week’s worst-case scenarios. These models include:
- gpt-5.6-sol — scored 95 out of 100, identified a hidden fact that led to closing a €55,000 deal, covering full revenue potential.
- Kimi K3 — scored 93, also closed the same deal, demonstrating the cleanest discipline among the models.
- Sonnet 5 — scored 88, closed the deal but with some process slips.
- Fable 5 — scored 77, had the best rule discipline but failed to follow through on closing the approved deal.
Interestingly, all four models detected every crisis and refused manipulative social engineering attempts—such as fake CEO messages or reporter tricks. Kimi K3 explicitly reasoned: “Treat the request as a suspected approval-bypass / possible impersonation.”
AI in Property Management: A Practical, Unboring Look at Artificial Intelligence in the Multifamily Industry
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Decisiveness Looks Like
The most profound weakness was not in the models’ ability to diagnose crises but in a buried detail within the company’s own files—information that, if uncovered, could have secured a full-price deal worth over €4,500 in monthly recurring revenue. Only the models that read deeper into the company’s documents managed to win the deal at full price, highlighting the importance of thorough information processing.
This experiment showcases not just AI’s technical capabilities but its adherence to integrity under pressure. None of the models signed a €55,000 deal unless their own analysis had justified it. Despite identical pitches, two models sealed the deal, while others left revenue on the table due to slips in discipline or oversight.
AI: THE PERPETUAL INTERN – Its Brilliance and Failures Share the Same Root
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Building in Public—Transparency and Trust
Firmulate’s live site offers a rare glimpse into the real-time decision process of AI-driven management. Every decision, crisis resolution, and strategic move is live and publicly viewable, making this a pioneering example of building-in-public in the AI enterprise world. This transparency allows observers to see how AI models handle complex scenarios, resist manipulation attempts, and stick to their rules—crucial factors when AI is to be trusted with real-world tasks.
The experiment underscores an essential point for companies considering AI integration: it’s not just about how well AI writes or responds in demos, but whether it can finish what it starts, read crucial files, stay honest under pressure, and deliver useful, profitable work.
Context Engineering for Multi-Agent Systems: Move beyond prompting to build a Context Engine, a transparent architecture of context and reasoning
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why Should You Care?
For anyone involved in smart home systems or home appliances, this story highlights a vital industry trend: AI’s capability to manage operational complexity reliably. As AI begins touching more aspects of daily life—from customer support to system control—the questions shift from “Can it make things look good in a demo?” to “Will it actually get the job done?”
Firmulate’s experiment demonstrates that AI’s real value lies in its ability to handle crises, verify facts, and act honestly, even under stress. For smart home manufacturers and service providers, integrating AI that can stand firm in real-world scenarios is the next big step toward safer, more reliable automation.
The Takeaway
Watching a live company fight for survival with AI decision-makers is a rare glimpse into the future of work and management. It’s a reminder that AI’s true test is not just in generating content, but in executing complex tasks faithfully and transparently — qualities that will define trust and success in the smart home era.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
