Imagine a world where artificial intelligence doesn’t just assist with customer support or data analysis but actually manages a company through its toughest week. How would different AI personalities handle crises, ethical dilemmas, and high-stakes negotiations? The answer is unfolding live, and it might challenge everything you think about AI and business management.
The Experiment: Putting AI to the Management Test
In a groundbreaking live experiment, four frontier AI models were tasked with running a small software company through its worst week. This real-world simulation involved identical crises, customer challenges, and temptations—yet each model’s decision-making style differed. Every decision was tracked, versioned, and auditable, providing a transparent look into how AI can handle complex management scenarios.
AI management decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Crisis Detection and Ethical Boundaries
Remarkably, all four AI models identified every crisis and refused every attempt at manipulation. In particular, they all rejected a staged social engineering attack involving fake CEO messages and a reporter trick. For instance, Kimi K3 explicitly treated such requests as possible impersonation, demonstrating a cautious and security-conscious personality.
As an affiliate, we earn on qualifying purchases.
The Critical Edge: Reading Between the Files
The decisive advantage came from reading company files deeply. Two models that examined information buried two document references deep in the company’s own files managed to identify a crucial detail—something that was invisible in surface-level analysis or chat interactions. This deep reading led to closing a full-price deal worth over €4,583 in monthly recurring revenue (MRR), significantly boosting the company’s performance.
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.
Decision Outcomes: Who Signed and Who Didn’t?
While all models recognized crises and refused manipulative tactics, only two ended up signing the €55,000 deal their analysis had earned. The other two, despite similar diagnoses and pitches, left the deal on the table. Interestingly, the models’ management personalities played a role: Opus 4.8, noted for its thoroughness and deep analysis, was last place because it slipped discipline and left opportunities unseized. Conversely, Kimi K3’s straightforward style resulted in the cleanest, most disciplined decision-making.
AI ethical decision support system
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Personality Profiles in AI
The experiment also highlights how different AI personalities can be measured: from the meticulous and cautious Opus 4.8 to the disciplined Kimi K3, and others with their unique traits. For example, all models refused social engineering attempts, but their approach to closing deals, reading files, and handling pressure varied significantly. This reveals that AI personalities are measurable and can be aligned with specific management styles.
The Bigger Implications: Trust, Efficiency, and Cost
The core takeaway is that the real test isn’t how well an AI writes or chats but whether it can complete tasks, stay honest under pressure, and read critical information before acting. In this live scenario, the best-performing models not only identified problems but also seized opportunities, demonstrating reliability and discipline vital for real-world decision-making.
The Platform: Wargaming AI Enterprises
Firmulate’s platform offers enterprises a way to simulate their AI workforce before deployment. By running these AI models against their own business data in a controlled environment, companies can assess management quality, ethical standards, and operational discipline. This live experiment offers a rare glimpse into how AI might perform in actual business settings, where real money and reputation are at stake.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
