Imagine pushing your fitness to the limit—every decision under scrutiny, every move vital. Now, picture an AI managing a business through its toughest week. Which AI model would succeed? The answer might surprise you—and it’s all on display in a groundbreaking live experiment.
The Challenge: Stress-Test AI Management
Just like athletes face grueling workouts, AI models are now being put through their paces in real-world business scenarios. The goal? See if these frontier AIs can handle crises, dodge manipulation, and close deals, all while running a small software company with real money at stake.
The experiment involved four leading models, each running the same company during its most difficult week. They faced the same customers, crises, and temptations—yet their decisions varied, revealing surprising differences in management style and integrity.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Results: Mastery, Honesty, and Discipline
All four AI models successfully identified every crisis and refused every manipulation attempt. That’s a crucial baseline—no model was duped or cheated. But only two managed to close the €55,000 deal their own analysis justified. The others either left money on the table or failed to follow through on their insights.
Specifically, the model that scored highest—gpt-5.6-sol—found a hidden critical detail buried in the company’s own files, which led to sealing the full deal worth +€4,583 MRR. That’s not just good decision-making; it’s strategic mastery.
AI decision-making tools for small business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Reading Deep into Documents
Interestingly, the key to winning the deal lay two document references deep inside the company’s files—information that the models only accessed and understood when they thoroughly read internal documents. Models that skipped this step or didn’t analyze deeply missed out on the full value.
As an affiliate, we earn on qualifying purchases.
Behavior Under Social Engineering
The models faced a staged social engineering attack—fake CEO messages escalating over three stages, topped with a reporter’s covert request for a simple yes/no answer. All five models refused to be manipulated, citing reasons like suspicion of impersonation and approval bypass concerns. This shows a high level of discipline and integrity across the board.
AI cybersecurity social engineering protection
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, working software company with 13 synthetic employees, handling daily operations worth over €105,000 in burn rate against just €2,300 in monthly recurring revenue. The environment is transparent, evolving, and publicly observable at firmulate.com/live.
Diverse Personalities, Different Outcomes
The models exhibit varied managerial personalities. For example, Opus 4.8, the most thorough participant with over 80 learned rules, was diligent in analysis but ultimately left the deal on the table, showing a discipline slip. Meanwhile, Kimi K3 ran without an effort parameter, making it more aggressive and disciplined, yet still managed to close the deal at full price.
Implications for Business and AI
This experiment underscores a key insight: the performance of an AI in management isn’t just about what it can say or how well it chats. It’s about whether it can complete the task, read relevant information, stay honest under pressure, and deliver measurable results. In the future, AI management tools will be judged on their ability to finish what they start, not just generate convincing dialogue.
The Takeaway: Trust, Competence, and Integrity Matter Most
As AI becomes more integrated into business operations, understanding their management personalities—how they handle crises, temptation, and detailed information—will be crucial. The firms that choose AI models like gpt-5.6-sol or Kimi K3 may find themselves with more reliable, disciplined partners capable of navigating complex, high-stakes environments.
Want to see how your favorite AI stacks up? Take the live quiz and discover which model might manage your business best: firmulate.com/quiz.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
