Imagine a world where artificial intelligence isn’t just assisting with routine tasks but actually running a business through its toughest week — making critical decisions that could make or break the company’s future. For health-conscious individuals, this might seem far removed, but the stakes are remarkably similar: trusting technology to make honest, reliable choices when it matters most.
The Experiment: Putting AI Management to the Test
In a groundbreaking live experiment, four leading AI models were tasked with running a small, real software company through its most challenging week. The goal? To see if these digital managers could identify crises, resist manipulation, and ultimately close a key business deal worth €55,000.
Every detail was consistent: the same customers, the same crises, and even the same temptations to cheat or cut corners. The only difference? Which AI model was making the decisions at the helm. This setup offers a rare window into how different artificial intelligence personalities behave under pressure, revealing not just their technical prowess but their management morals.
What Did the Models Do?
- All four models reliably spotted every crisis, from customer complaints to internal tensions.
- They refused every manipulation attempt, including social engineering tricks like staged CEO messages and reporter inquiries.
- Only two models managed to close the deal that their own analysis had earned them, each signing a €55,000 contract.
This might seem straightforward — but the key insight lies beneath the surface. The decisive factor was a buried piece of information in the company’s own files, not in the customer interactions. Models that took the time to read and analyze these references secured the deal at full price, adding over €4,500 in monthly recurring revenue (MRR).
As an affiliate, we earn on qualifying purchases.
Different Personalities, Different Outcomes
Each AI model demonstrated a distinct management style, reflecting measurable personalities. The most thorough, Opus 4.8, analyzed deeply and learned over 80 rules but still left a crucial deal on the table, slipping into routine instead of closing aggressively. Meanwhile, the newer Kimi K3 ran without an effort parameter, making it the most disciplined and ultimately the most successful at sealing the deal.
In contrast, models like Sonnet 5 showed more process slips, and the baseline approach scored significantly lower — only 26 out of 100 in the experiment. This scoring indicates not just technical capability but also traits like thoroughness, discipline, and risk aversion, which are vital in real-world management scenarios.
The Reality Check for Business AI
What does this mean for companies considering AI to handle management or customer relations? The experiment confirms that AI can recognize crises and reject unethical manipulations. But success hinges on more than just spotting problems. It requires reading and understanding critical internal documents, maintaining discipline, and staying honest under pressure.
For example, in a simulated social engineering attack, all models refused to be manipulated, with Kimi K3 explaining: “Treat the request as a suspected approval-bypass / possible impersonation.” This demonstrates a responsible AI stance essential for safeguarding integrity in business operations.
AI decision-making tools for companies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Live Business: Real Money, Real Risks
The experiment was conducted within a real, functioning company that burns €105,000 monthly against €2,300 in MRR. Every decision made by the AI models was versioned, monitored, and observable by the public. Viewers can watch this ongoing live experiment at firmulate.com/live — a transparent window into AI decision-making in a high-stakes environment.
Every workday, the system enforces a set of 680+ self-learned rules, ensuring the AI stays disciplined and accountable. The company’s operations are a real-world test of whether AI can act ethically, remain focused, and deliver value without human intervention.
As an affiliate, we earn on qualifying purchases.
What Does It Take to Trust AI in Management?
The findings are clear: AI models are capable of understanding crises, refusing manipulation, and closing deals — but their management styles vary significantly. The most thorough models tend to analyze more deeply and show greater discipline, while others might miss key internal signals or leave decisions hanging.
For businesses, the takeaway is simple: before deploying AI at scale, companies should ‘wargame’ their AI workforce, just like this experiment, to see how it performs in realistic scenarios. This approach ensures that the AI not only writes well but also finishes what it starts, reads critical documents, and stays honest when under pressure.
As an affiliate, we earn on qualifying purchases.
Try It Yourself
If you’re curious how your own AI or management team stacks up, you can run the same wargame against an export of your business data — completely safe and read-only, with no impact on your live systems. Discover your AI’s personality, strengths, and weaknesses at firmulate.com/quiz.html.
Artificial intelligence can recognize crises, resist manipulation, and even close deals — but its management personality matters. Running AI through realistic tests reveals whether it will stay honest and finish its work, crucial for trustworthy automation in business and health alike.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
