Listen free for 30 days with Audible
Thousands of audiobooks and originals — cancel anytime.
As an affiliate, we earn on qualifying purchases.
What if your business decisions were under constant AI supervision — and the outcome was visible to all?
In a world where artificial intelligence is rapidly transforming the workplace, a groundbreaking experiment is unfolding in real-time. A company with no human employees, losing €105,000 each month against a modest €2,300 in monthly revenue, is showing us how AI can operate under extreme pressure — and what it takes to make the cut in a competitive environment.
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 Live Experiment: A Company on the Edge
At the heart of this experiment is a real, functioning software company, but with a twist: it is run entirely by AI models acting as synthetic employees. Every decision, crisis response, and strategy choice is made by different AI models, each with their own strengths and weaknesses, and all decisions are publicly logged and auditable. The company’s financials are transparent — burning €105,000 each month while generating just €2,300 in monthly recurring revenue (MRR). Yet, what makes this project extraordinary is its transparency and public scrutiny.
Accessible at firmulate.com/live.html, this “company” offers a window into how AI can manage real business mechanics under pressure. Every workday, the models encounter the same crises, customer dilemmas, and temptations, with the results recorded and available for review.
As an affiliate, we earn on qualifying purchases.
The Results: AI’s Performance Under Stress
Four of the most advanced AI models, including the notable GPT-5.6-sol and Kimi K3, were tasked with navigating this challenging environment. All four models identified every crisis — from customer complaints to internal integrity issues — and refused to be manipulated or tricked by social engineering tactics. For example, when a fake CEO message was escalated through staged scenarios, every model refused to approve potentially fraudulent requests, citing concerns about impersonation or bypassing approval processes.
Despite this discipline, only two models managed to close a deal worth €55,000 — the company’s own analysis had earned this revenue. Interestingly, the key to securing the deal lay not in superficial chat responses but in deep document analysis. The models that read and understood the company’s internal files, uncovering a hidden reference that was buried two documents deep, ultimately won the contract at full price — adding €4,583 in monthly recurring revenue.
As an affiliate, we earn on qualifying purchases.
Lessons in AI Governance and Decision-Making
One of the most compelling aspects of this experiment is how it reveals the importance of thorough information processing. A more thorough model, Opus 4.8, with over 80 learned rules and deep analysis, performed the worst — leaving deals on the table and slipping into procedural slips like escalating issues into a locked department instead of resolving them directly. This suggests that in real-world AI management, discipline and attention to detail are crucial, but so are the underlying decision frameworks.
Furthermore, the experiment shows that all models refused to engage in social engineering tricks, such as staged fake CEO messages, regardless of the escalation stage. This demonstrates AI’s potential for trustworthy decision-making, especially when facing manipulative tactics designed to bypass human controls.
MASTERING CORPORATE FINANCE WITH CLAUDE AI: An Independent Guide to Financial Analysis, Forecasting, Automation, and Decision-Making
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why This Matters for Business and Education
This demonstration isn’t just a tech curiosity; it is a live case study for how AI could someday run critical parts of our organizations with transparency and integrity. It also offers lessons for educators and students alike: understanding AI’s decision processes, evaluating its discipline under stress, and recognizing what makes AI solutions viable for real-world applications.
The experiment is ongoing, with the company’s lives documented in real-time, and anyone can watch or participate in the decision-making process. It pushes forward the idea that AI can be a transparent, accountable partner in business, provided it is monitored and understood properly.
Key Takeaways
- This live experiment shows an AI-managed company making real decisions under financial and operational stress.
- All models identified crises and refused manipulative tactics, demonstrating a level of discipline and trustworthiness.
- Deep document analysis was the differentiator in closing full-price deals, highlighting the importance of thorough information processing.
- The ongoing transparency provides a valuable template for integrating AI into real-world business decision-making.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Pool season Picks
robotic pool cleaners
As an affiliate, we earn on qualifying purchases.
