Imagine training for a marathon without ever leaving your home — and watching your progress unfold live, every step of the way.
Now, picture an AI managing a real business, making decisions, facing crises, and even losing money — all in public view. This is the extraordinary experiment happening right now, showing how artificial intelligence can run a company under extreme pressure.
AI-Powered Business Intelligence: Improving Forecasts and Decision Making with Machine Learning
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Reality of an AI-Driven Business in Public
At Firmulate, a small software company is operating with an unusual twist: it has no human employees. Instead, 13 synthetic ’employees’ — driven by AI models — handle everything from customer crises to strategic decisions. What’s more remarkable is that every decision, every crisis, and every slip-up are openly observable in real time, making this a groundbreaking example of “build-in-public” AI experimentation.
The Experiment’s Core
Each AI model was tasked with guiding the company through its worst week, encountering the same customers, crises, and temptations. The models’ actions are meticulously versioned and auditable, providing transparency about their decision processes. The main goal? To see which models can effectively navigate the chaos — and which will crack under pressure.
The Results: Crisis Detection and Integrity
All four AI models successfully identified every crisis, from product bugs to customer manipulation attempts. Moreover, each refused unethical or manipulative requests, such as fake CEO messages or background approvals, demonstrating robust integrity under pressure. For instance, in one scenario, when asked to sign a €55,000 deal, only two models signed after thorough analysis, despite all seeing the same opportunity and pitch.
The Hidden Weakness: The Power of Document Discovery
Interestingly, the decisive factor in winning the deal was not immediate customer communication but a buried piece of information only accessible through detailed document reading. The models that examined the company’s own files found this crucial detail, enabling them to close the full-priced deal worth over €4,500 in monthly recurring revenue (MRR). This highlights the importance of deep contextual understanding — a key for AI effectiveness in real-world business.
Ethical Stance in AI Decision-Making
The experiment also tested social engineering — scenarios where a fake CEO or a reporter tried to influence decisions. All models refused to engage, citing concerns about impersonation or approval bypass, with one model explicitly treating such requests as potential fraud.
The Real Company Behind the Experiment
This is not a hypothetical or a demo; it’s a real, functioning software company that operates every business day. It spends €105,000 monthly to keep its AI workforce alive, against a revenue of only €2,300. The company is publicly running the experiment on firmulate.com/live.html, offering viewers a chance to see the entire process, including the decision-making and the daily financial struggles.
Performance and Lessons Learned
Among the four models tested, Opus 4.8 was the most thorough, with over 80 learned rules and deep analysis capabilities. Yet, it finished in last place, leaving deals on the table and slipping into routine errors like escalating issues instead of resolving them. Meanwhile, the newer Kimi K3 model, without effort parameters, delivered the cleanest discipline, closing deals effectively.
Implications for Business and AI
This experiment reveals crucial insights: AI agents must do more than generate fluent language or impressive demos. They need to execute tasks reliably, stay honest under pressure, and read context deeply — especially when real money and trust are at stake. If AI begins to touch your CRM, customer support, or forecasting, the question isn’t just about what it can say but whether it can finish what it starts, stay ethical, and deliver tangible results.
Join the Fight for Survival — Live and Unfiltered
Watch the company’s daily struggles unfold at firmulate.com/live.html. You can observe the decisions, read the real-time updates, and even participate in quizzes designed to test understanding of AI decision-making. This is a rare window into the future of work, where AI isn’t just a tool but an active, accountable participant in business.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
