Imagine managing a business under pressure, making critical decisions while facing crises and temptations to cut corners. Now, picture AI models stepping into that role, and the question arises: can they outperform human managers? Recent experiments with frontier AI technologies suggest they might, and the results are both fascinating and revealing.
Introducing the Live AI Management Wargame
At the heart of this investigation is a real, ongoing experiment conducted by Firmulate, an AI company that simulates entire companies running in real-time. Each simulation involves a small software firm navigating a week filled with customer crises, internal dilemmas, and ethical temptations—exactly the kind of stress tests that can expose whether an AI model can truly lead.
Four leading AI models, including GPT-5.6-SOL, Kimi K3, Sonnet 5, and Fable 5, were put through this intense management challenge. Every decision was recorded, and the same scenarios, crises, and manipulations were fed to each model to ensure fairness. The company’s real-world mechanics, with 13 synthetic employees and a cash flow of €105,000 monthly, provided a gritty backdrop that made the test both realistic and critical.
AI management decision support tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: AI Models Show Leadership and Integrity
Across the board, all four models demonstrated impressive vigilance—they spotted every crisis and refused every attempt to manipulate or cheat. For instance, when social engineering tactics like fake CEO messages and staged reporter requests emerged, all models refused to comply. Kimi K3’s explanation, for example, was that they treated such requests as potential impersonation attempts, highlighting a cautious, ethical stance.
The most striking outcome was that only two models managed to close a significant business deal worth €55,000. While all identified the core issues, only those two demonstrated the discipline to follow through and sign the deal, matching their own analysis and recommendations. The other two, despite understanding the situation, left the deal on the table, missing out on potential revenue.
business decision-making AI software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weaknesses—Reading and Acting on Internal Files
Interestingly, the decisive factor wasn’t just crisis detection but the models’ ability to access and interpret internal documents. The models that successfully secured the deal read two key references deep within the company’s files. That extra step—digging into company documents—proved crucial for making the right decision and capturing full revenue potential, worth more than €4,583 in monthly recurring revenue.
As an affiliate, we earn on qualifying purchases.
Understanding the Models’ Personalities and Decision Styles
Beyond raw decision-making, the experiment uncovered differences in management styles. The Opus 4.8 model, for example, was the most thorough, analyzing over 80 rules and providing the deepest assessments. Despite this, it was the least successful in executing the closing, leaving the deal unsealed and discipline slipping—highlighting that thoroughness alone doesn’t guarantee success.
Kimi K3, by contrast, ran without an effort parameter set at default, which contributed to its disciplined, straightforward approach, and ultimately, its success in closing the deal. These personality profiles hint that AI models can be characterized much like managers—some are meticulous, others are terse, and some are cautious or even hesitant under pressure.
As an affiliate, we earn on qualifying purchases.
The Real-World Significance
This live experiment isn’t just about testing AI in a laboratory setting. It reflects a broader question: if AI models are capable of making honest, effective decisions under pressure, they could revolutionize how businesses operate. Whether managing customer relationships, supporting complex supply chains, or making strategic choices, these models could offer a new level of reliability—if we understand their personalities and decision styles.
What sets this apart from standard AI demos is the transparency and real-world mechanics involved. The company’s operations, with its daily decision logs, self-learned rules, and live performance, are open for observation. Anyone can watch the process unfold at firmulate.com/live and see how each model performs, in real time.
Why This Matters for Your Business and Wellness
For readers interested in health and wellness, the takeaway extends beyond business: understanding AI’s decision-making capacity is akin to understanding how your body or mind responds under stress. Just as your health depends on your resilience and honesty, a business’s success may depend on how well its AI systems stay transparent and disciplined under pressure.
Whether you’re considering AI for customer service, internal management, or strategic planning, the key questions are: does it finish what it starts? Does it read relevant information thoroughly? Does it stay honest under stress? The Firmulate live experiment offers clear evidence that AI models can meet these standards—and that their personalities, like human managers, can be characterized and chosen accordingly.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
