Can AI Managers Be Trusted? Real Decisions, Real Models, Real Stakes

AIThis post was created with the assistance of artificial intelligence (AI).

Imagine running an interior design firm where every decision counts—choosing suppliers, managing client expectations, and staying honest under pressure. Now, picture having AI models act as your virtual managers, making critical choices in your stead. But how do you know which AI is reliable? Are they all the same, or do some have personalities that could affect your business’s integrity? Welcome to a groundbreaking experiment where real AI models are put through their paces, revealing not just their decision-making skills but their managerial personalities.

At Firmulate, a company dedicated to testing AI’s business acumen, four advanced frontier AI models are put through a simulated week of crisis—mirroring the worst week an interior design or furniture firm might face. The goal? See if these AI ‘managers’ can handle real-world challenges, maintain honesty, and close deals, just like human managers.

Each AI model operates in the same environment, dealing with identical customer requests, supply chain issues, and ethical dilemmas. Every decision is recorded and auditable, ensuring transparency in their reasoning. The models are scored based on their ability to spot hidden information, refuse manipulative tactics, and ultimately secure a meaningful contract.

The Results Are Eye-Opening

All four AI models identified every crisis and refused every attempt at deception—showing they can recognize critical issues and uphold integrity.
Only two models managed to close a deal worth €55,000, aligning with their own analysis and decision-making. The other two either hesitated or left money on the table despite their correct diagnosis.
Interestingly, the decisive advantage came from a subtle detail buried two documents deep in the company’s files—the models that read this information won the full deal, worth an additional €4,583 in monthly recurring revenue.

This experiment underscores an important point: the decision-making personality of AI models matters. While they all recognize crises, their willingness to follow through and close deals depends on their internal ‘personality’.

Personality Profiles in AI Management

The experiment also sheds light on different AI ‘personalities’. For example, the most thorough participant, Opus 4.8, analyzed over 80 rules and provided detailed insights. Yet, it left the deal unclosed, showing a disciplined but cautious personality. Conversely, other models displayed more straightforward, terse approaches or refused noise communication altogether, reflecting different managerial styles.

Engineering Trust in AI

In social engineering tests, all models refused to approve suspicious requests or background questions from fake reporters, demonstrating their capacity to resist manipulation. Kimi K3, for instance, explained: “Treat the request as a suspected approval-bypass / possible impersonation.” This consistency across models highlights that, at a fundamental level, AI decision-making can be aligned with ethical standards—if designed accordingly.

The real-world application of this experiment is clear. Firms deploying AI to handle customer relations, project bids, or supply management need models that don’t just write convincingly—they must finish what they start, read relevant files thoroughly, and stay honest under pressure.

The Live Business Simulation

The experiment runs in a real software environment with 13 synthetic employees managing a small, money-losing firm. They burn €105,000 monthly against €2,300 in monthly revenue, illustrating how AI decisions impact real cash flow. Every workday, the firm’s processes are versioned, making the results transparent and measurable. Watch the live decision-making at firmulate.com/live.

What This Means for Interior Design and Decor

For interior designers and furniture retailers considering AI tools, the takeaway is clear: the quality of decisions matters as much as language fluency. An AI that recognizes crises, reads from relevant files, and follows through can be a trustworthy partner. Conversely, models that leave deals on the table or fail to read crucial details could jeopardize your projects and reputation.

Before hiring AI for your business, consider running a wargame—like the one at firmulate.com/quiz.html—to see how different models perform in simulated crises. This way, you can select an AI that embodies the managerial personality you need: diligent, honest, and decisive.

The Future of AI in Business

This experiment suggests that AI models can have measurable personalities, influencing their management style and trustworthiness. As AI becomes more integrated into client interactions, project management, and strategic decisions, understanding these personalities will be crucial to maintaining integrity and delivering results.

So, next time you think about automating your project bids or client communications, remember: it’s not just about language. It’s about choosing an AI that can finish what it starts, read the files it needs, and stay honest under pressure—just like a good manager.