Imagine a real company losing €105,000 every month, yet you can watch its management decisions unfold in real time — all while AI models compete to steer it through its worst week. This is not a simulation; it’s the live experiment conducted by Firmulate, revealing how AI manages crisis, honesty, and strategic decision-making in a high-stakes environment.
The Live Business Experiment: A Company Without Employees
At the heart of this story is a small, real software company running every business day with a twist: it has no human employees. Instead, it’s operated by 13 synthetic “employees,” guided by over 680 self-learned rules, and subject to constant versioning and public scrutiny. The company’s financials are stark — it burns through €105,000 each month while generating only €2,300 in monthly revenue. Yet, despite the losses, this experiment offers an unprecedented look into how AI models handle real-world management decisions under pressure.
AI Co-Thinking: A Framework for Working with AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
AI Models in the Crisis Hot Seat
Four frontier AI models, including the well-known gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8, were tasked with navigating the same challenging week for this company. Every decision they made was recorded, time-stamped, and auditable, creating a transparent and ongoing story of how AI responds to crises, manipulative attempts, and ethical dilemmas.
The Critical Findings
- All four models identified every crisis the company faced, from customer issues to operational challenges.
- They refused every manipulation or social engineering attempt, including fake CEO messages and reporter tricks.
- Only two models managed to close a €55,000 deal, which their own analysis earned — a key indicator of strategic competence.
- The decisive weakness was linked to a buried reference in the company’s internal files, which the models that read those files could exploit to win the deal at full price, worth over €4,500 in monthly recurring revenue.
Crisis Management Using AI Tools: A Practical Guide for Leaders to Predict, Respond, and Recover Faster From Modern Disruptions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Insights Beyond the Demos
This experiment reveals that AI’s ability to detect hidden opportunities — like a crucial detail buried in internal documents — can be a game-changer in real business negotiations. Interestingly, while models like Opus 4.8 performed the most thorough analysis, they also slipped on discipline, leaving deals on the table or failing to escalate issues internally. This highlights that even the most advanced AI can struggle with processes and discipline, not just decision-making.
Responsible AI: Implement an Ethical Approach in your Organization
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Human and Ethical Dimension
The models also faced social engineering tests, such as staged CEO messages and background inquiries. All five models refused to cooperate, reasoning that such requests could be impersonation or approval bypass attempts. This demonstrates that AI can be trained to uphold integrity and ethical boundaries, even under pressure — a vital trait for future business tools.
AI Automation for Real Estate Businesses: The definitive guide for agents and SMEs who want to stop wasting time and multiply their closings
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why This Matters for Your Business
What does all this mean for companies that plan to integrate AI into their management, CRM, or customer support? The key takeaway is that it’s not just about whether an AI can generate convincing chat responses. It’s whether the AI can finish what it starts, read and analyze internal files, stay honest under pressure, and deliver measurable work. These capabilities determine whether AI can truly replace or augment human management in complex, high-stakes situations.
The League Table: Who Performs Best?
- gpt-5.6-sol: 95 points — found the buried fact, closed the deal, and demonstrated complete performance.
- Kimi K3: 93 points — closed the deal too, with the cleanest discipline of the field.
- Sonnet 5: 88 points — managed to close the deal but with some process slips.
- Opus 4.8: 77 points — also closed the deal but left some opportunities on the table due to discipline issues.
Watch and Learn
The live experiment is fully accessible at firmulate.com/live. You can watch the company run every business day, explore the decisions made, and see firsthand how AI models manage crises, ethics, and strategic negotiations in real time. Additionally, the experiment includes a quiz where you can guess which model made which decision, based on 242 real, unedited management choices — a fascinating way to see AI decision-making in action.
Build in Public — The Extreme Version
This is the most transparent and public AI management experiment ever attempted. It’s a raw, unfiltered look at what AI can and cannot do in a real company with real money, real crises, and real consequences. For enterprises considering AI tools, the takeaway is clear: success depends not just on how well the AI writes or responds, but on its ability to finish tasks, read internal information, and uphold honesty under pressure.
This unprecedented live experiment shows AI’s potential and limits in real management. The key: AI must read, finish, and stay honest — or risk losing deals and trust.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
