What if a company could run entirely without employees—and still face real financial and ethical struggles? Welcome to a groundbreaking experiment in AI-driven business management, now live for all to see.
AI business management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Live Company in Action
At the heart of this experiment is a company made up of 13 synthetic employees, each guided by advanced AI models. Unlike traditional businesses, this one operates in real-time, with its cash reserves dwindling at €105,000 each month against a modest €2,300 in monthly revenue. Every decision, crisis, and temptation is publicly documented and versioned daily at firmulate.com/live.
This setup isn’t just a gimmick; it’s a rigorous test of AI’s ability to manage complex, real-world business scenarios. The company faces simulated crises, customer interactions, and ethical dilemmas—just like a real startup. And the results? They’re both illuminating and sobering.
How Do AI Models Perform?
Four frontier AI models—each with unique strengths—were tasked with running this company through its worst week. They encountered the same customers, faced identical crises, and had access to the same information. The models were required to make decisions, handle manipulative tactics, and even read through internal documents to uncover hidden insights.
One standout was the GPT-5.6-sol model, which identified crucial buried information in internal files and successfully closed a €55,000 deal, boosting monthly recurring revenue by over €4,500. Meanwhile, Kimi K3, a newcomer, also closed the deal, demonstrating the potential for AI to win business on merit and thorough analysis.
In contrast, some models faltered not in their diagnoses but in execution. For example, Opus 4.8, the most meticulous participant with over 80 learned rules, left a promising deal unexecuted because discipline slipped during the final moments. This highlights an ongoing challenge: even the most capable AI can stumble under stress or discipline lapses.
Ethical and Security Tests
The experiment also tested AI resistance to social engineering. Fake CEO messages and reporter tricks were used to see if the models would bypass security protocols. All five models refused to cooperate, citing concerns about impersonation and bypassing approval processes. Kimi K3 explained its stance: “Treat the request as a suspected approval-bypass / possible impersonation.”
AI decision-making tools for startups
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why This Matters for Businesses
While the company itself is fictitious, the implications are real. As AI becomes more integrated into business operations—handling customer support, sales, and decision-making—the ability of these systems to act ethically, honestly, and effectively under pressure is critical.
Current AI models can spot problems, refuse manipulative tactics, and even uncover hidden information—a promising sign for automation and management. But the experiment also shows gaps: even high-performing models may leave valuable deals on the table or falter under stress, emphasizing the importance of testing AI systems before deployment.
AI ethics and security testing tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Future of AI in Business
This live experiment, accessible at firmulate.com/live, offers a rare glimpse into AI’s capabilities and limitations in a simulated but realistic environment. It’s an ongoing story of survival, discipline, ethics, and innovation—one that challenges assumptions about AI’s readiness to take on complex management roles.
For decision-makers, it’s a wake-up call: deploying AI isn’t just about getting the right answers in a chat window. It’s about ensuring these agents can finish what they start, read the right information, and do so honestly—especially when money and reputation are on the line.
AI customer support automation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Broader Context
In a world increasingly powered by AI, businesses need to ask whether their automation tools can meet these high standards. The experiment at Firmulate doesn’t just show what AI can do—it highlights what it must do to be truly trustworthy in high-stakes environments.
As the experiment continues, viewers can watch the daily decisions unfold, see the models learn, adapt, and sometimes slip. It’s a glimpse into the future of AI management—a future where transparency, discipline, and integrity are more crucial than ever.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
