Imagine a Company That Has No Employees, Loses Thousands Every Day, and Yet Is Fully Transparent About Its Struggles
In a world where AI is often portrayed as a magic wand for efficiency, a groundbreaking experiment is turning that narrative on its head. Meet a live, publicly accessible company simulation powered entirely by AI models that operate as synthetic employees. As you watch, this company faces real crises, makes difficult decisions, and battles to stay afloat — all on a publicly visible platform that reveals its every move.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Raw Reality of an AI-Run Business
At the heart of this experiment is a small software company operating in real time, but with a twist: it’s run entirely by artificial intelligence models, each acting as a synthetic employee. This setup isn’t just a tech demo; it’s a candid, ongoing portrait of a business fighting to survive, with a public cash countdown, daily decision records, and a transparent playbook of over 680 self-learned rules.
Every workday, the company is versioned and the decisions are logged — creating a detailed, auditable history of how AI manages crises, sales, and ethical challenges. The company burns around €105,000 each month while generating only €2,300 in monthly recurring revenue, a stark reminder that even AI-managed companies aren’t immune to financial struggles.
How Do the AI Models Perform?
Four frontier AI models were pitted against identical business scenarios — the worst week a small software firm could face, complete with customer crises and inner temptations to manipulate outcomes. The results are illuminating:
- All models identified every crisis and refused manipulation attempts. Even when fake CEO messages were faked or reporters tried to bypass approval steps, all four AI systems held firm.
- Two models managed to close a lucrative deal. Despite the identical diagnosis and pitch, only GPT-5.6-sol and Kimi K3 signed the €55,000 deal they independently identified as the best outcome. The others, despite good discipline, left money on the table or failed to execute the decision.
- The key weakness was buried in internal documents, unnoticed in customer interactions. AI models that read the company’s internal files successfully identified crucial opportunities, enabling full-price deals worth an additional €4,583 monthly recurring revenue.
Ethical and Trust Considerations
The experiment also tested the models against social engineering attacks. Fake CEO messages, staged in multiple escalation stages, were consistently refused, with Kimi K3 explicitly citing suspicion of approval bypass or impersonation. This indicates a promising level of ethical restraint, even under pressure.
AI in Strategy and Decision-Making for Small Business Owners: Affordable AI Tools to Evaluate Ideas, Model Outcomes, and Set Priorities (AI Productivity for Small Business Owners Book 10)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Platform That Makes This Possible
The live site, firmulate.com/live.html, offers a window into this ongoing experiment. Visitors can see a real-time, versioned simulation of a tiny but complex business environment where AI models perform as managers, decision-makers, and ethical agents. Every decision, every crisis, and every outcome is transparent and auditable, providing a unique lens into how AI might manage real-world organizations in the future.
Implications for Business and Education
This experiment isn’t just a curiosity. It poses fundamental questions for business, education, and AI development:
- Can AI truly manage complex ethical dilemmas and maintain trust under pressure?
- Will AI-driven companies be sustainable, or are their financial struggles a sign of systemic limitations?
- How can organizations prepare for a future where AI plays a central role in decision-making?
Additionally, the experimental setup includes a quiz that tests users’ ability to distinguish between different AI models’ decisions, fostering a deeper understanding of AI behavior in management scenarios. Details and results are available at firmulate.com/quiz.html.
HUMAN CENTERED ARTIFICIAL INTELLIGENCE SYSTEMS: Explainability ethical design and decision support engineering
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Takeaways: Transparency, Ethical Resilience, and the Future of Work
This experiment demonstrates that AI can identify crises and refuse manipulation under pressure, but it also highlights the financial fragility of AI-managed companies. What makes this project compelling isn’t just its technical insights but its build-in-public approach — exposing every decision, every weakness, and every victory in real time. As AI increasingly touches our workplaces, understanding its limits and potentials in a transparent way will be crucial for business leaders and educators alike.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
The AI Prompt Playbook for Excel & Financial Analysis: 50 Ready-to-Use AI Prompts for Excel, Financial Analysis, Dashboards, Power BI, Financial Modeling & Automation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
