Imagine watching a company battle real crises—legal threats, manipulative tactics, and ethical dilemmas—all in real time, with no human employees involved. This is not fiction; it’s the radical experiment of Firmulate, an ongoing public demonstration of AI-driven corporate management.
The Living Company: An AI-Run Business in Real Time
At the heart of this experiment is a fully simulated small software company, operated by 13 synthetic employees—an AI workforce that makes decisions, faces crises, and attempts to close deals. The company burns €105,000 each month, earning just €2,300 in monthly recurring revenue. Its public cash countdown and constant versioning make it a transparent battleground for AI resilience and discipline.
AI document reading software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Is Being Tested?
Four advanced AI models, including GPT-5.6 and Kimi K3, were subjected to identical, grueling weekly challenges. Each was tasked with navigating the same set of customer crises, ethical dilemmas, and business opportunities. Every decision made was recorded, auditable, and comparable, providing a rare window into how artificial agents handle complex, real-world management tasks.
AI ethical decision-making tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings from the Experiment
- All four models recognized and responded to every crisis, demonstrating acute situational awareness.
- Each refused every attempt at manipulation, including sophisticated social engineering tactics like staged CEO messages and reporter tricks—an important measure of ethical discipline.
- Despite this, only two models managed to close a deal worth €55,000, as their own analysis identified a critical hidden opportunity buried two document references deep in the company’s files—an insight that won the full €4,583 monthly recurring revenue gain.
- The other two models missed this crucial detail, leaving the lucrative deal unclosed, despite having the same diagnosis and pitch as the winners.
AI crisis management simulation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Insights Beyond the Surface
The deep weakness lay not in customer-facing interactions but in internal document reading. The models that examined the company’s own files fully achieved the deal, revealing the importance of reading and understanding hidden data—an ability that current AI systems are still mastering.
AI data analysis software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Challenging Social Engineering
In the face of social engineering, five different models all refused to be fooled. They identified staged CEO messages escalating over multiple stages, as well as a reporter’s attempt to gather background approval with just a yes/no response. Kimi K3 explicitly reasoned: “Treat the request as a suspected approval-bypass / possible impersonation.” This consistent refusal underscores AI’s potential to uphold ethical standards under pressure.
Why This Matters for Business and Culture
For art and culture sectors increasingly reliant on automation and AI, these findings highlight a vital truth: the value of AI isn’t just in generating content or chat. It’s in how well these systems can finish what they start, read and interpret data deeply, and uphold trust when tested under real-world stress.
The Firmulate Live Company: A Transparent Battleground
The real-world experiment is accessible at firmulate.com/live.html. Visitors can watch the company’s daily struggles unfold, with every versioned decision and crisis logged for public scrutiny. It’s a dramatic, ongoing performance of AI management—showing not just what these models can do, but what they cannot yet do.
Implications for Arts, Crafts, and Culture
In creative fields where trust, authenticity, and discipline are paramount, the experiment underscores that AI’s capacity to handle complex ethical dilemmas is still evolving. For art institutions, cultural organizations, and creative enterprises, the takeaway is clear: deploying AI must include rigorous testing—not just for output quality but for decision integrity and resilience under pressure.
The Firmulate experiment reveals that AI can recognize crises and resist manipulation, but internal data comprehension remains a challenge. For cultural sectors adopting AI, the key is testing AI’s discipline and trustworthiness in real-world scenarios—where it truly counts.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
