Can AI Manage a Business Under Pressure? Watch and Decide
Imagine a company with no human employees, yet one that faces real crises, makes tough decisions, and struggles to stay afloat—all in public view. This is not fiction. It’s the live experiment by Firmulate, where artificial intelligence models operate a small software business in real time, battling to survive against the odds. For educators, scientists, and anyone curious about the future of AI in decision-making, this ongoing story offers unprecedented insight into how machines handle the complexities of real-world management.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Experiment: AI as a Business Manager
Firmulate’s live demonstration runs four frontier AI models as the entire management team of a tiny software company. Each model, from GPT-5.6 to Sonnet 5, faces identical challenges: same customers, same crises, same temptations to cheat or manipulate. Every decision—whether dealing with crises, negotiations, or ethical dilemmas—is versioned and auditable, creating a transparent record of how each AI navigates a difficult week in business.
This setup is extreme but revealing. The models are tested not merely on chat-based performance but on their ability to manage a company’s real mechanics—cash flow, trust, decision discipline—and to do so honestly under pressure. The company burns €105,000 each month, with only €2,300 in monthly recurring revenue, and is publicly counting down its cash. All decisions made by the models are publicly accessible at firmulate.com/live.html.
The AI-Driven Leader: Harnessing AI to Make Faster, Smarter Decisions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What the Data Shows: Competence and Integrity
The results are striking. All four models identified every crisis as it arose and refused all attempts at manipulation, including complex social engineering tactics like fake CEO messages and reporter tricks. For instance, when fake requests escalated or involved impersonation, all models refused to budge—Kimi K3 explicitly recognized the risk of approval bypass or impersonation, reinforcing their hesitance.
However, at the critical moment, only two models managed to close the deal that their analysis had earned—the €55,000 contract. The other two, despite diagnosing the opportunity correctly, left the deal on the table or failed to follow through, demonstrating the importance of discipline and decision execution under pressure.
HUMAN CENTERED ARTIFICIAL INTELLIGENCE SYSTEMS: Explainability ethical design and decision support engineering
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Read Your Files
Beyond surface decisions, a deeper weakness emerged. The models that read the company’s internal files and references deeply—specifically, the buried information in internal documents—were able to identify a crucial overlooked opportunity. This allowed them to win a deal at full price, adding over €4,583 in monthly recurring revenue. Meanwhile, models that did not delve into these internal references missed this opportunity entirely.
An Introduction to Healthcare Informatics: Building Data-Driven Tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Implications for Business and Education
This experiment is more than a corporate stunt. It reveals that AI’s capacity to handle complex, high-stakes decisions depends on its ability to read and interpret relevant internal information, resist manipulation, and execute disciplined actions. For educators and researchers, it highlights the importance of transparency and verifiable decision-making in AI systems—skills essential for trust and reliability.
Moreover, the live setup exposes the gaps and weaknesses in current models, emphasizing the need for thorough testing before deploying AI in sensitive roles like management, support, or strategic planning. As the experiment continues, it offers a rare window into how AI might someday manage real-world tasks—if it can be trusted to stay honest and finish what it starts.
What You Can Observe and Learn
Anyone interested in the future of AI, whether educators, students, or business leaders, can watch this experiment unfold at firmulate.com/live.html. The platform provides a real-time view of decision-making processes, decision versions, and the ongoing performance of different AI models facing the same operational stress test.
This transparent approach allows viewers to understand that AI’s readiness is not solely about generating convincing chat. It’s about executing complex management tasks, resisting manipulation, and acting with discipline—skills critical for AI to be truly trustworthy in the workplace.
Key Takeaways
- AI models can identify crises and refuse manipulation in real-time management scenarios.
- Deep internal knowledge—like reading company files—is crucial for recognizing hidden opportunities and winning deals.
- Discipline and decision execution are as vital as diagnosis, especially in high-pressure environments.
- The live experiment offers a rare view into how AI manages real-world business mechanics, providing lessons for education and industry alike.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
