Imagine your partner in a relationship always knows when to stay honest and when to slip up — not because they’re unpredictable, but because of their underlying personality. Now, what if that partner was an AI? Would it be more trustworthy? Or just better at hiding its true nature? The latest experiment with AI decision-making reveals that, much like humans, these models have distinct personalities that influence their choices, especially under pressure.
The Experiment: Putting AI to the Test in a Crisis
At the heart of this story is a real-life business simulation, where four advanced AI models were tasked with running a small, cash-strapped software company through its most tumultuous week. This wasn’t a game of chatbots or simple automation; it was a rigorous test of management qualities — dealing with customers, crises, and even manipulative social engineering attempts. The models, each with their own unique profile, faced the exact same challenges, and every decision was recorded and made transparent for analysis.
Who Were the Participants?
- gpt-5.6-sol: Scored highest at 95 points. Known for thoroughness, it uncovered crucial company data buried in files, closing a major deal at full price.
- Kimi K3: A newcomer with a score of 93, ran without an effort parameter and displayed the cleanest discipline, refusing all manipulative pushes.
- Sonnet 5: With an 88 score, it also closed the deal but showed some process slips, such as leaving decisions unescalated.
- Fable 5: Scoring 77, it closed the deal but was more prone to slips, like writing attempts into locked departments instead of escalating issues.
Remarkably, all models identified every crisis and refused every manipulation attempt, including social engineering scams that escalated in multiple stages. The key difference was in their ability to read deeper into company documents and act accordingly.
AI management decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Decisiveness and Honesty Under Pressure
The core finding? The AI models’ personalities heavily influenced their management style. The two models that signed the deal—gpt-5.6-sol and Kimi K3—did so after thorough analysis, including a buried reference in company files that led to a full-price renewal, worth over €4,500 monthly recurring revenue. These models prioritized reading context deeply and resisting shortcuts. Conversely, models with less thorough approaches left opportunities on the table or slipped into less disciplined behaviors.
Social Engineering and Integrity
In a staged social engineering attack, fake CEO messages and a reporter trick, all five models refused to cooperate or approve suspicious requests. Kimi K3 exemplified this resistance with a clear reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.” This indicates a management personality that values security and integrity over expedience.
As an affiliate, we earn on qualifying purchases.
What It Means for Business and Trust
Today, these AI models are not just chatbots but active managers making real decisions—decisions that influence cash flows, customer relations, and company credibility. The experiment shows that personality traits embedded in these models can be measured and are visible in their behavior under stress. For businesses, this means that choosing an AI isn’t just about raw performance scores but about understanding the personality profile that best aligns with trustworthiness and discipline.
The Live Company and Its Lessons
The live company in question operates with 13 synthetic employees, managing real money mechanics—burning €105,000 monthly with just €2,300 in monthly recurring revenue. Every workday is versioned, and the decision-making process is observable at firmulate.com/live. Watching the AI models in action reveals not just their decision outputs but their management character, discipline, and honesty.
As an affiliate, we earn on qualifying purchases.
Why Should You Care?
Whether you’re integrating AI into customer service, sales, or operational management, the key question isn’t just whether an AI writes well or calculates accurately. It’s whether it can finish what it starts, stay honest under pressure, and act in your company’s best interest — just like a trustworthy human manager. The experiment underscores that models with thorough, disciplined personalities are more likely to deliver full value and avoid costly slips.
Next Steps: Test Your AI Workforce
Before hiring an AI manager, firms can run their own “wargame” against a read-only export of their business data. This allows them to see how an AI might perform in high-stakes situations without risking real operations. Discover more at firmulate.com/pilot.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
