Imagine facing a crisis at your company—customers demanding answers, files buried with crucial info, and temptation to cut corners. Now, picture passing that test not once, but across multiple AI decision-makers, each with a distinct personality. How would your AI team perform? The answer may surprise you.
Testing AI’s Management Skills in Real Time
In a groundbreaking live experiment, four advanced AI models were tasked with running a small, real software company through its worst week. This wasn’t a staged demo; it was real-world business mechanics—customers, crises, temptations, and money flowing in real-time. Every decision was recorded, every crisis was the same across models, and none could cheat or manipulate the system undetected.
AI decision-making software for business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Models and Their Personalities
- gpt-5.6-sol 95: The top performer, who not only identified hidden information but also secured the €55,000 deal on their own analysis.
- Kimi K3 93: A newcomer with the cleanest discipline, also closed the deal, but with slightly fewer process slips.
- Sonnet 5 88: Managed to close the deal but demonstrated some slips in process discipline.
- Fable 5 77: Also closed, yet with more process slips, and left some opportunities on the table.
As an affiliate, we earn on qualifying purchases.
Key Findings: Honesty, Insight, and Decision-Making
Despite their different personalities, all models successfully identified crises and refused manipulative attempts, such as social engineering tricks. For example, they refused fake CEO messages escalating over three stages and even a reporter’s manipulative request, reasoning: “Treat the request as a suspected approval-bypass / possible impersonation.”
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness: Reading Deeper Files
The decisive factor wasn’t in the obvious customer crises but buried two document references deep in the company’s files. The model that read beyond the surface found the information needed to close the deal at full price—adding €4,583 MRR to the company’s pipeline. This underscores a critical point: in management, often what’s unseen is what determines success.
As an affiliate, we earn on qualifying purchases.
Real Business Mechanics at Play
The experiment mimicked a running company with 13 synthetic employees, managing real money—burning €105,000 monthly against €2,300 MRR. The environment was transparent and watchable at firmulate.com/live, where every decision and outcome unfolds in real-time. The company operates under 680+ self-learned rules, with daily versions of its work processes, emulating the complexities of real management.
Different Personalities, Different Outcomes
The most thorough participant, Opus 4.8, with over 80 learned rules and deep analyses, finished last—leaving deals on the table and slipping into departmental silos instead of escalating crucial issues. Interestingly, all models exhibited similar weaknesses, suggesting that even the most advanced models can struggle with certain discipline areas under pressure.
Implications for Business and AI Integration
This test isn’t just about AI capabilities; it’s about management integrity and decision quality. If AI agents will someday handle your customer relations, forecasts, or strategic decisions, it’s vital to ask: Will they finish what they start? Will they read and understand your internal files? Can they resist manipulation and stay honest when it counts? The experiment shows that some AI models can do all of this—and some, surprisingly, can’t.
Try It Yourself
Business leaders interested in assessing their AI’s readiness can run similar tests against their own data, safely and without risking real damage. Visit firmulate.com/quiz.html for a live, interactive quiz built on real decisions and watch how different AI models perform in controlled management scenarios.
The Bottom Line
While AI chat demos often impress with language, the real test is whether they can manage operations reliably under pressure. The models’ ability to identify hidden insights, resist manipulation, and complete critical tasks determines if they’re ready for prime time. As AI continues to integrate into business, understanding their management personalities could be as important as their technical prowess.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
