Imagine a business operating without employees, yet constantly struggling to stay afloat — and all of it visible in real time. Welcome to the world of Firmulate, a groundbreaking experiment in build-in-public AI management, where an AI-driven company fights for survival every single day.
The Live Experiment: Running a Company in Real Time
Firmulate has created a fully operational, publicly accessible company simulation. It features 13 synthetic employees, each guided by a complex set of over 680 self-learned rules. The company’s mechanics are real: it burns €105,000 every month against a tiny €2,300 monthly recurring revenue (MRR). This setup isn’t just a demo; it’s a real-time experiment, visible at firmulate.com/live, where anyone can watch the company’s daily struggles unfold.
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 Core of the Experiment: Testing AI Decision-Making Under Pressure
Four advanced AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—are each tasked with managing the same week of crises, customer interactions, and temptations. The goal? To see how well they can navigate real-world business challenges without human intervention. Each decision is versioned and kept fully auditable, ensuring transparency in how the AI responds to every crisis and opportunity.
Excel Simulations
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Surprising Results: Spotting Crises and Maintaining Integrity
- All four models identified every crisis that arose during the week.
- All refused every manipulation attempt, including social engineering tactics like fake CEO messages and reporter tricks.
- Only two models managed to close a €55,000 deal, earning full marks for diagnosis and pitch — but surprisingly, neither signed the deal in the end.
The critical insight emerged from an overlooked document in the company’s files: the model that read this information and incorporated it into its decision-making was the one that successfully signed the deal at full price, adding +€4,583 in monthly revenue.
Generative AI Security: Theories and Practices (Future of Business and Finance)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What This Means for Business and AI
This experiment underscores a key point often missed in many AI demos: the ability to identify what truly influences decision-making. The models that read and interpret internal documents — not just external cues — perform significantly better at closing deals and maintaining integrity. For cloud and infrastructure providers, this highlights the importance of comprehensive data access and trustworthy AI integration in operational decision-making.
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.
Handling Social Engineering and Ethical Challenges
The company also tested the AI’s response to social engineering. Over three escalating stages, a fake CEO message and a reporter’s background request were presented. All models refused to be manipulated, with Kimi K3 explicitly treating the requests as potential impersonation or approval bypass attempts. This demonstrates the growing capacity of AI to uphold ethical standards even under pressure.
The Reality of Managing a Money-Losing Business
Firmulate’s live setup isn’t glamorous. It’s a small-scale company burning €105,000 monthly while generating just €2,300 in recurring revenue. Yet, it’s a vital testbed for understanding how AI can help manage such dire situations, or at least provide insights into decision quality under extreme conditions.
Deep Dive into the Models’ Performance
Among the participants, Opus 4.8 — the most thorough with over 80 learned rules and the deepest analysis — finished last. It left a deal unexecuted due to a discipline slip, showing that more rules and analysis don’t always translate into better outcomes. Kimi K3, running without an effort parameter, showed the clearest discipline and signed the deal at full price, illustrating that even simpler configurations can outperform more complex ones under certain conditions.
Implications for the Future of AI in Business
This experiment is a glimpse into a future where AI models are not just chatbots but operational managers capable of handling real crises, reading internal documents, and maintaining ethical standards under pressure. For companies in cloud, hosting, and infrastructure, it raises critical questions:
- Will your AI agents be able to finish what they start?
- Can they read and interpret your internal data securely?
- Will they stay honest under stressful conditions?
The ongoing live experiment at firmulate.com/live offers a rare window into the capabilities and limitations of AI-driven management. It’s a test of trust, discipline, and transparency—qualities that will define the future of AI in enterprise operations.
Firmulate’s live AI management experiment reveals that reading internal documents and resisting manipulation are crucial for trustworthy AI in business. Watch it unfold at firmulate.com/live.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
