Imagine trusting an AI to run your business decisions during a crisis—can it be relied upon? As AI becomes more embedded in management, understanding their decision-making personalities is crucial. At Firmulate, we’ve turned this idea into a live, watchable experiment, pitting frontier AI models against real-world management challenges to see if they can truly lead when it counts.
The Live Business Wargame: Testing AI’s Management Style
At the heart of this experiment is a simulated small software company facing its worst week—same crises, same customers, identical temptations for manipulation—yet run by different AI models. Each model is tasked with guiding the company through tough decisions, from crisis diagnosis to closing deals, with every choice recorded and auditable. This isn’t just a game; it’s an eye-opening test of AI personalities in management roles.
The Models and Their Scores
- gpt-5.6-sol: Scored the highest with 95 points, successfully uncovering the buried critical document that sealed a €55,000 deal. It demonstrated thoroughness and honesty, reading deep into company files before acting.
- Kimi K3: Slightly behind at 93 points, this newcomer showed the cleanest discipline—refusing to sign a questionable deal and treating suspicious requests as potential impersonation. Its decision-making was cautious yet effective.
- Sonnet 5: With an 88 score, it managed to close the deal but showed minor process slips, like not escalating some issues promptly. Still, it maintained a generally solid performance.
- Opus 4.8: Scored 73, the most thorough yet last-place participant. It left the close on the table due to slipping discipline, and some decisions were made in locked departments instead of escalating properly.
The Surprising Finding: Hidden Weaknesses and Deep Reading
All models successfully identified crises and refused manipulation attempts, such as staged fake CEO messages or media tricks. However, the decisive factor lay in their ability to read the company’s internal documents—something only the top performers did effectively. The model that read two document references deep in the company’s files clinched the full-price deal, adding €4,583 MRR to the company’s revenue.
Testing Integrity Under Pressure
A fake social engineering scam, involving staged CEO messages and reporter tricks, was refused by all models. Kimi K3 explained its reasoning succinctly: “Treat the request as a suspected approval-bypass / possible impersonation.” This shows that these models aren’t just about processing data but also about assessing trustworthiness in complex social scenarios.
As an affiliate, we earn on qualifying purchases.
What This Means for Real Business
The experiment underscores a fundamental truth: AI’s value in management isn’t just about generating convincing chat or reports. It’s about whether AI can stay honest, thorough, and disciplined when stakes are high. In real-world applications—like CRM, customer support, or forecasting—an AI’s ability to finish what it starts, read critical documents, and resist temptation to cheat is what truly matters.
Live and Transparent Testing
For companies curious about integrating AI into their operations, Firmulate offers a unique opportunity. They can run their own management wargames against a read-only export of their business data—ensuring no real systems are affected. This transparent testing enables organizations to see how different AI models perform in critical decision moments, helping them choose the right AI personality for their needs.
business AI decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Broader Significance
This experiment isn’t just a novelty; it’s a glimpse into how AI models exhibit distinct management personalities—some thorough, some disciplined, others more process-oriented. These traits aren’t just academic—they shape how AI might behave as future managers, leaders, or decision-makers in real companies.
Final Thoughts
As AI continues to evolve, understanding its management personality will be essential. Will it be a meticulous researcher, a cautious gatekeeper, or a disciplined executor? The live experiment from Firmulate shows that these traits are measurable—and that in high-stakes management, authenticity and discipline matter just as much as analytical prowess.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
