Imagine running a restaurant or a bustling kitchen — every decision matters, from handling an upset diner to managing inventory shortages. Now, picture an AI taking charge in this chaotic environment, making critical choices during its worst week. Would it stay honest, decisive, and effective? This is exactly the question a groundbreaking live experiment by Firmulate explores, and the results might surprise you.
Get kitchen staples and gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
How AI Models Face the Heat of Real Business Crises
In a unique live experiment, four frontier AI models were tasked with managing a small software company during its most turbulent week. Just like a chef facing a full house, these models had to handle the same customers, crises, and temptations — but each one ran a different version of management style. The goal? Measure not just their smarts, but their integrity, decisiveness, and consistency under pressure.
What Was Tested?
- Recognizing and responding to crises
- Refusing manipulative or unethical requests
- Closing deals and managing company finances
Every decision was recorded and made transparent, creating a real-time showcase of AI decision-making in a high-stakes environment. The models faced a series of tests, including:
- Customer complaints and service crises
- Attempts at manipulation, like fake CEO messages and reporter tricks
- Financial negotiations and deal closures
The Key Findings
All four models successfully identified every crisis and refused every unethical manipulation, demonstrating robust integrity. But when it came to finalizing the most profitable deal, only two models managed to sign the €55,000 contract earned through their own analysis. The other two, despite diagnosing the issues correctly, left money on the table due to discipline lapses.
This reveals a critical nuance: in the heat of crisis management, it’s not enough to spot problems or refuse manipulation — decisive action and commitment to follow through are equally vital. Interestingly, the most successful model, GPT-5.6-sol, was also the one that uncovered hidden information deep inside the company’s files, leading to a full-price deal worth an additional €4,583 in monthly recurring revenue.
Products Worth Considering
KitchenAid Classic Series 4.5 Quart Tilt-Head Stand Mixer K45SS, Onyx Black
Cuisinart SM-50BK Stand Mixer, 5.5 Quart, 12 Speed, Onyx
Cuisinart 8-in-1 Air Fryer Toaster Oven with Grill, 1800W, Stainless Steel
As an affiliate, we earn on qualifying purchases.
The Hidden Weakness in AI Decision-Making
One of the most revealing aspects of the experiment was the importance of thorough information reading. The model that read two document references deep into the company’s files secured a lucrative deal, while those that didn’t missed out on the full potential. It underscores that in real business, understanding context deeply can make or break outcomes — a lesson as relevant to chefs as to AI managers.
Social Engineering Tests and Ethical Stance
The models faced staged social engineering attempts, including escalating fake CEO messages and a reporter trick asking for a quick yes/no on background. All models refused to cooperate, with Kimi K3 explaining, “Treat the request as a suspected approval-bypass / possible impersonation.” This highlights an essential trait: ethical AI should recognize and resist manipulative tactics, maintaining integrity even under pressure.
Real Business, Real Money, Real AI
The experiment was conducted on a live, functioning company with 13 synthetic employees, operating real money mechanics — burning €105,000 monthly against €2,300 in monthly recurring revenue. Every workday, the company’s management decisions are versioned and observable at firmulate.com/live. This setup offers a rare glimpse into how AI models perform in an authentic business environment, not just in canned demo scenarios.
The Management Personalities Revealed
The models showcased distinct personalities:
- GPT-5.6-sol was the most thorough and decisive, uncovering hidden truths and closing the deal at full price.
- Kimi K3 was the newcomer, demonstrating the cleanest discipline and closing the deal too, but with slightly less depth.
- Sonnet 5 showed solid performance, but with more slips in process and discipline.
Meanwhile, Opus 4.8, the most comprehensive in analysis, failed to close the deal, leaving money on the table due to discipline lapses—like writing attempts into a locked department instead of escalating them properly.
What Does This Mean for Your Business?
Are you considering AI to manage customer relationships, support, or forecasts? The key takeaway is that performance isn’t just about writing well or generating ideas; it’s about completing tasks, reading deeply into your data, and maintaining honesty under pressure. The models with the highest scores in the experiment also demonstrated the strongest management traits.
Test Your AI Readiness
If you want to see how your AI workforce might perform, you can run your own wargame against a read-only export of your business. This allows you to measure decision-making quality without risking your actual operations. Find out more at firmulate.com/pilot.html.
Final Thoughts
The live experiment by Firmulate proves that AI’s management personality matters just as much as its intelligence. In real-world scenarios, trustworthiness, decisiveness, and thorough understanding can unlock hidden value and prevent costly mistakes. As AI begins to touch more facets of your business, ask not just if it can write well — but if it can finish what it starts and stay honest under pressure.
Real AI management isn’t just about intelligence; it’s about trust, decisiveness, and thoroughness. Firmulate’s live experiment shows that only the most disciplined models close the deal and unlock hidden value—crucial traits for AI in business.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
