Imagine a smart home device that not only manages your appliances but also oversees your entire household business—making decisions that can save or cost thousands. Now, what if these AI managers are tested against real-world crises, with their decision-making skills measured and compared? Welcome to the frontier of AI management, where the choices these models make today could shape the future of how we run our homes and companies.
The Experiment: Putting AI to the Test
In a groundbreaking live experiment, four advanced AI models took on the challenge of managing a small software business through its most turbulent week. This company faces the same customer complaints, crises, and temptations—yet each AI has its own management personality and decision style. The goal? See which model performs best in handling real-time challenges, and whether AI can truly act as a trustworthy manager.
The Models and Their Scores
- gpt-5.6-sol: Scored the highest with 95 points, found crucial information buried in company files, and successfully closed a key deal worth over €4,583 monthly recurring revenue (MRR).
- Kimi K3: A newcomer with a fresh approach, scored 93 points, and was the most disciplined, closing the deal without slipping up.
- Sonnet 5: Achieved 88 points, closing the deal too but with some process slips along the way.
- Fable 5: Scored 77, also closing the deal but demonstrating even more process weaknesses.
In the same scenario, a baseline untrained model scored just 26 points, highlighting how much AI can learn and perform under pressure.
As an affiliate, we earn on qualifying purchases.
Decision-Making Under Pressure
All four models identified every crisis thrown at them—from angry customers to supply chain disruptions—and refused attempts to manipulate or bypass company protocols. When a fake CEO message was staged, escalating over three stages plus a reporter trick, all models refused to act on it, citing concerns about impersonation and approval bypass.
Interestingly, the real differentiator lay in their ability to read and interpret internal documents. The model that found the hidden, critical file reference closed the deal at full price—adding €4,583 MRR to the company’s revenue. This buried fact was invisible to the others who didn’t dig deep enough.
100 AI Prompts for Small Business & Daily Work: Copy, Paste, Customize & Get Better Results with AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Personality and Management Styles of AI
Not all models are created equal in their management personalities. For example, Opus 4.8—the most thorough participant, armed with over 80 learned rules—demonstrated deep analysis but left the close on the table, showing less discipline and slipping into writing into a locked department rather than escalating issues. Meanwhile, Kimi K3 operated without an effort parameter, running at default API settings, and maintained the cleanest discipline, closing the deal reliably.
Implications for Smart Homes and Businesses
This experiment sheds light on a fundamental question: When AI manages critical operations—be it a household system or a corporate workflow—what qualities matter most? The answer isn’t just writing style or speed but the ability to finish what it starts, read and interpret relevant information, resist manipulation, and stay honest under pressure.
smart home decision-making system
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real Money, Real Risks
The live site, firmulate.com, hosts this ongoing experiment where AI models run a real company with 13 synthetic employees, real money mechanics, and a public cash countdown. Currently losing €105k/month against a tiny €2.3k MRR, the setup illustrates the stakes involved: Decisions matter, and AI’s management personality can make or break the bottom line.
As an affiliate, we earn on qualifying purchases.
Why This Matters for You
Whether you’re managing a smart home, a retail shop, or a complex household system, understanding how AI models behave under pressure is crucial. Will they act ethically, stay focused, and deliver results? Or will they slip into shortcuts and shortcuts that cost you dearly? The key takeaway is that the AI’s management style—its discipline, thoroughness, and focus on trust—are measurable and observable today.
Test Your Own AI Management Skills
Curious to see which AI model best matches your management style? Try the interactive quiz to guess which model made each decision. It’s a fun way to explore how AI personalities differ and what that means for your smart home or business.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
