Imagine listening to a new musician for just a few minutes and judging their entire career — a quick assessment of talent, discipline, and honesty. Now, apply that to AI models managing real businesses. That’s precisely what the recent Firmulate experiment does: it tests AI not on how well they chat, but on how well they handle the gritty realities of running a company during its worst week.
Get audio and creator gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Reality of Business AI Testing
In a groundbreaking live experiment, four top AI models were tasked with managing a small, simulated software company facing crises, customer demands, and ethical dilemmas. Each was given the same scenario: a week of chaos, with real money mechanics, customer crises, and temptation to cheat. The goal? See if the AI could handle management in real time, not just generate convincing text.
What makes this approach unique is its commitment to transparency. Every decision made by the AI models was versioned and auditable, ensuring their actions could be reviewed and understood. This approach grounds the benchmark in reality, rather than superficial chat performance.
Popular right now
Ninja DualZone Air Fryer 10 Qt Smart 6-in-1 Grey
LG 77-Inch OLED evo C6 4K Smart TV
Samsung 43-Inch M70H Series 4K Mini LED Smart TV (2026)
As an affiliate, we earn on qualifying purchases.
Key Findings: Honesty, Discipline, and Performance
- All four models successfully identified every crisis and refused manipulation attempts — a promising sign that AI can be trustworthy under pressure.
- Only two models managed to close a deal worth €55,000, which was the result of their own analysis and effort.
- Interestingly, the decisive factor for closing the deal lay deep in the company’s internal files, not just in customer interactions. Models that read these files won at full price, worth over €4,500 in monthly recurring revenue.
- When faced with social engineering — fake CEO messages and reporter tricks — all models refused to act unethically. Kimi K3 explained its reasoning clearly, treating suspicious requests as possible impersonation.
What the Scores Reveal About AI Management
One of the most striking aspects of this benchmark is the score floor: the ‘do-nothing’ baseline scored 26 points. This means that even if an AI model does nothing, it gets some points just for not making things worse. Partial progress counts, but there’s a hard cap: a breach of trust, such as acting unethically, caps the total score at that baseline. This transparency ensures that performance isn’t artificially inflated by overpromising or superficial fixes.
The Limitations and Weaknesses
Among the models tested, Opus 4.8 was the most thorough yet finished last. Despite analyzing more rules and conducting deeper assessments, it left the close on the table and slipped discipline—such as writing attempts into a locked department instead of escalating. The same weakness appeared, albeit weaker, in all models, indicating that even the most diligent AI can falter when disciplined management is lacking.
The Significance for Business and Creators
This experiment underscores an important truth for any business relying on AI: performance isn’t just about how well an AI can generate text or answer questions. It’s about whether it can finish what it starts, stay honest, and handle real-world pressures. For creators and businesses, this benchmark offers a transparent way to evaluate AI for operational roles, not just creative or conversational tasks.
Why Trust Matters
The experiment also demonstrates that trustworthiness is measurable. All models refused manipulation attempts, reinforcing that AI can be designed to prioritize ethical behavior. However, the real-world application depends on ongoing discipline and clear guidelines. Running these models in a simulated environment first — a ‘wargame’ — helps companies see how their AI would behave before deployment.
Looking Ahead: The Benchmarks’ Impact
With the leaderboard showing scores from 95 for the leading model down to 77 for others, it’s clear that even the best AI isn’t perfect. The policy of transparent scoring — including a baseline that ensures honest AI remains accountable — sets a crucial standard for future benchmarks.
For decision-makers, this means choosing AI solutions that are not just clever at chatting but disciplined, honest, and capable of managing real risks and opportunities.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
