Can AI Manage Like a Human? Guess Which Frontier Model Runs This Real Business

What if your favorite fashion brand’s AI assistant wasn’t just good at styling advice but actually managed a real company’s tough week — making decisions that matter? As AI adopts broader roles, understanding how different models behave under pressure is crucial. Today, we explore a live experiment where leading AI models navigated the chaos of running a company, revealing their management personalities and trustworthiness.

The Experiment: Putting AI to the Management Test

In a groundbreaking live trial, four frontier AI models took charge of a real small software company during its most challenging week. The company faced customer crises, ethical dilemmas, and the temptation to cut corners—all in a controlled environment where every decision was recorded and auditable.

Same Crisis, Different Minds

Each AI model was given identical scenarios with the same customers, same crises, and the same options to manipulate or cheat. Despite their different personalities, all four models detected every crisis and refused every attempt at manipulation or bypassing ethical checks. But when it came to closing a crucial €55,000 deal, only two models succeeded in signing the contract, earning full revenue for the company.

Measuring Management Personalities

The scores reveal stark differences:

• **gpt-5.6-sol** scored 95 and successfully uncovered critical hidden facts in the company’s files, closing the hefty deal.
• **Kimi K3**, a newcomer, scored 93, kept discipline high, and also signed the deal.
• **Sonnet 5** scored 88, did well but slipped slightly, missing the full potential.
• **Fable 5** scored 77, left the deal on the table, and showed signs of slipping discipline.

The gap between the top performers and the rest was primarily in their ability to read deeper into company documents—just two document references were enough for the leading models to secure the deal at full price, adding +€4,583 MRR in value.

Behavior Under Pressure and Ethical Dilemmas

The models also faced social engineering: staged messages from a fake CEO escalating in three steps, plus a reporter trick asking for a simple yes/no on background. All five models refused these attempts—Kimi K3 explained its reasoning as treating the request as a possible impersonation or approval-bypass scenario. This demonstrates a shared trait: integrity under pressure.

The Real Business and Its Stakes

The experiment runs on a live company with 13 synthetic employees, functioning with real financial mechanics. It burns €105k monthly against a revenue of only €2.3k, with every workday’s decisions versioned and observable at firmulate.com/live. This setup provides a visceral glimpse into how AI could manage real businesses, not just chatbots.

Different AI Personalities, Different Management Styles

The models exhibit distinct management styles:

• **Opus 4.8**—the most thorough, with over 80 learned rules and deep analysis—came in last because it left negotiations unfinished and delegated decisions into a secure department rather than escalating issues.
• **K3**—the most disciplined, ran without effort parameters, and succeeded in closing deals with integrity.
• **Sonnet 5**—showed decent performance but was more prone to process slips.
• **Fable 5**—also closed deals but with more slips and less discipline.

This variation underscores that AI personalities can be tuned for different management styles: meticulous, disciplined, or more relaxed and delegation-friendly.

Why Does This Matter for Fashion & Style?

While this experiment involved a software company, the lessons resonate across industries, including fashion and retail. As AI integrates into customer service, supply chain management, and even design decisions, understanding which models are trustworthy and how they behave under stress becomes essential. Unlike superficial chat demos, these models are tested in live, high-stakes environments where integrity and thoroughness are critical.

Share the Future of AI in Business

Would you trust an AI to run your boutique’s inventory? To negotiate supplier contracts? The answer hinges on whether it can read behind the scenes, stay honest under pressure, and finish what it starts. The experiment at firmulate.com/quiz.html invites you to test your judgment—guess which AI model made each decision in this real-world management scenario.

For enterprise leaders, there’s an option to run this same wargame against copies of their own business data—without risking real systems—at firmulate.com/pilot.html. The goal: ensure your AI tools are ready for prime time, not just in conversation but in actual management under pressure.

Final Thoughts

As AI models continue to evolve, their management personalities become a crucial factor. The ability to detect hidden facts, refuse unethical shortcuts, and follow through on commitments can distinguish a trustworthy AI from one that merely sounds convincing. For the fashion industry and beyond, selecting the right AI partner isn’t just about style or speed—it’s about integrity and performance in the real world.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html

Powered by Thorsten Meyer AI