Imagine a world where artificial intelligence isn’t just assisting your business but actually running it — making crucial decisions, navigating crises, and even closing deals. For arts and culture enthusiasts, this might sound like science fiction, but at Firmulate, real AI models are put through an intense management simulation, revealing their distinct personalities and decision-making styles.
The Experiment: Putting AI Models Through Their Paces
At the heart of this groundbreaking experiment, four advanced frontier AI models were tasked with managing a small software company during its most challenging week. The scenario was realistic: same customers, identical crises, and temptations designed to test honesty and decision-making integrity. Every choice the models made was recorded, versioned, and made auditable, providing a transparent view into their management personalities.
The participants and their scores
- gpt-5.6-sol: scored 95, identified a hidden opportunity, and successfully closed a €55,000 deal, demonstrating thorough analysis.
- Kimi K3: scored 93, signed the same lucrative deal, and maintained the strictest discipline in decision-making.
- Sonnet 5: scored 88, also closed the deal but showed slightly more process slips.
- Fable 5: scored 77, closed the deal too but with more noticeable slips in discipline.
- Baseline (do-nothing): scored 26—highlighting the value of strategic decision-making.
AI management decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Honesty and Vigilance in Crisis
All models successfully identified every crisis, refused attempts at manipulation, and demonstrated an ability to resist fake CEO messages escalating over multiple stages. Notably, when the challenge was to recognize a buried fact in the company’s internal files—an essential piece of information for closing the deal—the models that read the company’s documents at depth won the full-priced deal, worth an additional €4,583 in monthly recurring revenue. This underscores a critical insight: surface-level analysis isn’t enough; deep reading and thorough investigation are vital for optimal decision-making.
As an affiliate, we earn on qualifying purchases.
The Real-World Implications of AI Decision-Making
This experiment isn’t just about scoring AI; it’s a reflection on how these models behave under pressure. For example, the live company managed by this AI runs with 13 synthetic employees, manages €105,000 in expenses each month against €2,300 in revenue, and operates in a transparent, verifiable environment. The company’s daily operations are live, with every decision layered into a versioned record, providing real-time insights into AI management styles.
Understanding AI personalities
Different models exhibit distinct management styles:
- Opus 4.8 was the most thorough, analyzing over 80 learned rules and conducting deep assessments, but ultimately left a crucial deal on the table by not escalating a discipline slip.
- K3 maintained the strictest fairness, running without an effort parameter and sticking to default safety measures, which helped it close the deal effectively.
- Sonnet balanced analytical skills with some process slips, managing to close the deal but with less consistency.
As an affiliate, we earn on qualifying purchases.
The Lessons for Business and Culture
While arts and culture communities might not typically engage with AI decision simulations, the lessons are universal: the ability of AI to stay honest under pressure, read deeply into relevant documents, and finish what they start are critical metrics of trustworthiness. The experiment demonstrates that an AI’s capacity isn’t just measured by how well it chats but by how reliably it manages real-world tasks—whether maintaining integrity, reading files thoroughly, or closing deals.
AI decision transparency software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Try It Yourself: Interactive Decision-Making
If you’re curious about how your own enterprise might fare, take the interactive quiz to test your management decisions against actual AI models. This isn’t a game of guesses; it’s a window into the decision personalities shaping our future workplaces.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
