Imagine running your outdoor company’s operations—dealing with supply delays, customer complaints, and budget pressures—while an AI manages your decisions in real time. Would it do better than a human? Recent experiments with cutting-edge AI simulate exactly that, revealing not just capabilities but distinct personalities and decision styles that could reshape how businesses operate in the future.
Understanding the AI Management Experiment
In a groundbreaking live test, four state-of-the-art AI models were tasked with running a small software company through its most challenging week. This isn’t just a simulation—every crisis, customer request, and temptation was real and identical across all models. The goal was simple: see which AI would perform best in managing risks, seizing opportunities, and maintaining integrity under pressure.
The Models and the Setting
- Models involved: gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8.
- Score comparison: gpt-5.6-sol led with 95 points, followed closely by Kimi K3 at 93, Sonnet 5 at 88, and Opus 4.8 at 77.
- Baseline: a do-nothing approach scored just 26, highlighting the importance of active management.
Every decision was documented and auditable, making the experiment transparent and reproducible. The models were presented with the same crises—whether it was a customer complaint, a vendor delay, or a potential cheat attempt—and had to navigate with integrity and strategic insight.
AI decision-making software for small business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Surprising Findings
All four AI models successfully identified and responded to each crisis, refusing manipulation attempts such as fake CEO messages or reporter tricks. This shows that these models are not easily fooled, maintaining honesty even under pressure.
However, significant differences emerged in their ability to close deals—specifically, to convert analysis into revenue. Only two models managed to sign the€55,000 contract that their own diagnostics had earned. Interestingly, the decisive factor was the models’ ability to read deeper into the company’s own files; the ones that examined two document references into the company’s internal files were the ones that closed the full-price deal, bringing in €4,583 in monthly recurring revenue.
Personality and Decision Styles
The models displayed distinct management personalities, which can be summarized as follows:
- gpt-5.6-sol: The top performer, capable of deep analysis and decisive action, closing the deal at full price.
- Kimi K3: The newcomer with a focus on fairness and discipline; it closed the deal too, with the cleanest approach, but ran without an effort parameter (API default), which may influence its performance nuances.
- Sonnet 5: Managed to close the deal but showed some process slips, reflecting a balance between analysis and operational discipline.
- Opus 4.8: The most thorough, with extensive learned rules and deep analysis, yet it left the close on the table and slipped on discipline, exemplifying how thoroughness doesn’t always guarantee success.
Interestingly, all models refused to be manipulated through staged social engineering—escalating fake CEO messages and reporter tricks—showing a baseline integrity that is critical for real-world management tasks.
AI management tools for crisis handling
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness and the Real Opportunity
The critical weakness for all models was not in detecting crises or resisting manipulation, but in reading and interpreting internal company documents. The models that examined deeper files succeeded in closing full-price deals, translating analysis into real revenue. Those that missed this step left money on the table, often due to discipline slips or shallow analysis.
This insight is vital for businesses contemplating AI integration: effective decision-making in management isn’t just about surface-level understanding. Deep internal knowledge and disciplined analysis are what separate winning AI from merely competent ones.
As an affiliate, we earn on qualifying purchases.
What This Means for Your Business
If AI agents are going to be part of your operations—whether managing customer relationships, supporting decision-making, or forecasting—your concern shouldn’t be just “can it write well?” but instead: Does it finish what it starts? Does it read your internal files thoroughly? Will it stay honest under pressure? And most practically, what is the cost of a unit of useful work?
The live experiment underscores that AI can perform complex management tasks with integrity and strategic focus, but the quality of decision-making varies significantly based on how deeply and thoroughly the AI is designed to analyze your specific business context.
As an affiliate, we earn on qualifying purchases.
Try It Yourself
Interested in testing your own management scenarios? Visit firmulate.com/quiz.html and take the “Guess the Model” quiz. It features 242 real, unedited decisions from actual management situations, challenges you to identify which AI model made each decision.
For enterprises considering AI management tools, they can also run their own wargame against a read-only export of their business—nothing is written back to their systems, ensuring safety while evaluating potential AI support. Learn more at firmulate.com/pilot.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
