Imagine a beauty brand trying to keep its promises under scrutiny—not just from customers but from the AI systems managing its operations. What if these AI agents, designed to run the business, faced real crises, temptations, and pressures just like human managers? And what if you could watch the entire experiment unfold in real time? That’s exactly what a groundbreaking live experiment by Firmulate is revealing about the future of AI in business.
Open a free Amazon Business account
Business pricing, bulk buying and tax-exempt orders.
As an affiliate, we earn on qualifying purchases.
The Live Company That Runs on AI—Yet Loses Money Every Day
At the heart of this experiment is a real, functioning software company managed entirely by AI models. Unlike typical demos, this company is not just simulating tasks—it’s actively running, day by day, with 13 synthetic employees making decisions that cost real money. Currently, it burns €105,000 each month while generating only €2,300 in monthly recurring revenue, creating a dramatic live cash countdown.
Every decision made by the AI is meticulously tracked, versioned, and publicly accessible at firmulate.com/live.html. This transparency allows anyone to see how AI models handle crises, temptations, and complex management decisions in real time. The company’s purpose? To assess whether AI can not only strategize but also execute reliably, honestly, and effectively under pressure.
As an affiliate, we earn on qualifying purchases.
Testing AI’s Business Integrity in Extreme Conditions
In a recent test, four advanced AI models were challenged to manage the same small software company through its worst week. They faced identical crises, same customer issues, and temptations to cut corners or manipulate data. The goal was simple yet revealing: could the AI identify every crisis, refuse manipulation attempts, and secure a lucrative deal?
Results showed all models correctly recognized each crisis and refused every manipulation. Yet, only two managed to close the deal worth €55,000 in monthly recurring revenue, aligned with their own analysis. The other two also identified the opportunity but failed to sign the contract, highlighting an intriguing gap: crucial competitive insights lay hidden in the company’s internal documents, not in the surface-level customer interactions.
As an affiliate, we earn on qualifying purchases.
What’s Behind the Deal—And Why It Matters
The winner, GPT-5.6-sol, uncovered a buried fact deep within the company’s files, giving it a decisive edge. This demonstrates that in real-world AI management, success often hinges on reading and understanding internal data—not just reacting to external crises. It’s a reminder that AI’s ability to process information thoroughly can be a game-changer in high-stakes business environments.
As an affiliate, we earn on qualifying purchases.
Human-Like Temptations and AI’s Ethical Boundaries
Another revealing component of the experiment involved social engineering attacks—fake messages from a CEO escalating over three stages and a reporter’s subtle background request. All four models refused to bypass security or approve dubious requests, with Kimi K3 explicitly reasoning that the request was a suspected impersonation. This indicates a growing maturity in AI’s capacity to adhere to ethical boundaries, even under simulated pressure.
As an affiliate, we earn on qualifying purchases.
The Reality of a Build-in-Public Experiment
This isn’t just a theoretical test. The company operates in real-time, with daily decision-making and live updates at firmulate.com/live.html. Every workday, the company’s decisions are versioned and open for scrutiny, offering a transparent view into how AI models handle complex management tasks. It’s a build-in-public experiment that reveals the strengths and weaknesses of AI-driven decision-making in a real business context.
Lessons for the Beauty and Personal Care Industries
While this experiment involves a software business, its implications resonate across industries like beauty and personal care. The core question is: can AI systems, which might soon manage customer relations, inventory, or marketing, be trusted to finish what they start, stay honest under pressure, and read critical internal data?
For brands that rely on trust—whether it’s in product claims, customer engagement, or supply chain integrity—the ability of AI to perform reliably is paramount. Watching this live company struggle and succeed in real-time offers a sobering perspective on what’s possible—and what’s still challenging—with AI automation.
The Road Ahead: Building Trust in AI’s Business Capabilities
As AI models continue to evolve, so too must our understanding of their real-world performance. The experiment by Firmulate illustrates that success isn’t just about how well AI can chat or generate content, but whether it can handle crises, avoid manipulation, and uphold business integrity—especially when money, reputation, and trust are on the line.
For decision-makers in every industry, the message is clear: test your AI systems in the trenches, not just in demos. The live experiment at firmulate.com/live.html offers a rare window into the future of trustworthy AI in business—one where transparency and resilience are non-negotiable.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Flea & tick season Picks
flea and tick prevention
As an affiliate, we earn on qualifying purchases.
