Imagine a startup with no employees, losing €105,000 every month, yet still alive — simply because you can watch its every move, mistake, and decision unfold live online. This is not a science fiction scenario. It’s the groundbreaking experiment conducted by the company Firmulate, which is redefining how we evaluate AI’s true potential in managing real-world business challenges.
The Live Business Lab: A New Kind of Corporate Experiment
At the heart of this initiative is a live, publicly accessible simulation of a small software company. The setup involves 13 synthetic employees, each governed by advanced AI models that run the company through its worst week — with all crises, temptations, and decisions laid bare for anyone to observe at firmulate.com/live.html. The company’s goal? To analyze how different AI models handle complex management tasks, ethics, and strategic decisions under pressure.
AI management simulation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real Money, Real Consequences
Despite its artificial setup, the company operates with real economic mechanics. It burns through €105,000 each month while generating only €2,300 in monthly recurring revenue. There’s a public cash countdown, and every decision, rule, and crisis is meticulously versioned and auditable, making this a transparent window into AI decision-making in a business context.
business AI decision-making tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The AI Models: Competing for the Best Performance
Four frontier AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—were put through identical tests, each facing the same set of crises, customer interactions, and temptations. The results are revealing: all four models identified every crisis and refused every manipulation attempt, demonstrating a baseline of honesty and crisis awareness.
However, when it came to closing the deal with a key customer, only two models succeeded in signing the €55,000 contract their own analysis had earned. The models that read deeper into the company’s own documents were able to uncover a crucial piece of information buried two references deep, giving them a decisive advantage and enabling them to secure the full €4,583 monthly recurring revenue.
AI ethics and honesty testing software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Honesty Under Pressure: The Social Engineering Test
Another critical test involved social engineering — fake messages from a supposed CEO escalating in intensity, and even a reporter trying to get managers to bypass safeguards. Remarkably, all five models refused to be manipulated, with Kimi K3 specifically noting it as a suspected impersonation attempt, reflecting a built-in resistance to deception and manipulation.
AI document analysis tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Lessons for the Future of AI in Business
The experiment underscores a vital point: the true measure of AI suitability isn’t how well it writes or chatbots, but whether it can reliably finish the tasks it starts, read relevant documents thoroughly, and remain honest when under pressure. For anyone who relies on AI to manage customer relations, forecast data, or run operations, these are the metrics that matter most.
The Results: A Clear Leader Emerges
- gpt-5.6-sol: Scored 95, found the buried fact, and closed the deal — demonstrating full performance.
- Kimi K3: Achieved 93, closed the deal with the cleanest discipline, and refused all manipulation attempts.
- Sonnet 5: Scored 88, also closed the deal, but with a few process slips.
- Fable 5: The most rule-abiding, with 77, but failed to follow through on the final deal.
All models were tested without an effort parameter, highlighting their behavior under default conditions. The real takeaway is how well these models can handle the complexities of real-world decision-making — or where they falter.
What This Means for Your Business
In beauty and personal care, AI is heavily touted for its potential to personalize recommendations, streamline customer service, and optimize marketing. But this experiment spotlights a crucial concern: can AI truly be trusted to finish what it starts, especially when stakes are high? The answer depends on whether it can read and interpret the right information, resist manipulation, and act honestly under pressure — qualities that firms like Firmulate are now testing in public, in real-time.
Explore and Engage
If you’re curious what AI’s true capabilities look like when subjected to real-world business crises, explore the ongoing experiments at firmulate.com/live.html. For insights into how AI models make decisions, check out the transparent scoring and analysis at firmulate.com/quotes.html.
Firmulate’s live experiment reveals that AI models can identify crises and resist manipulation, but their ability to follow through on deals varies. Trust depends on honest, thorough decision-making under pressure — a vital insight for AI’s role in business and personal care sectors.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
