Imagine a business that operates entirely without employees, yet is perpetually losing money — and you can actually watch its daily struggles unfold in real time. This is not science fiction but a groundbreaking experiment in the world of artificial intelligence and management.
The Live Experiment: A Company Without People
At the heart of this bold venture is a publicly accessible online platform that simulates a small software company run entirely by AI models. Known as the live experiment, it features 13 synthetic employees, each powered by cutting-edge AI models, navigating the chaos of real crises, customer demands, and business decisions — all recorded and versioned daily.
AI business management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unparalleled Transparency and Reality
The company burns through €105,000 every month, earning just €2,300 in monthly recurring revenue, with a public cash countdown highlighting its precarious financial state. Every decision made by the AI workforce is logged, auditable, and openly displayed, offering a rare glimpse into how these models perform under pressure in a live setting. This transparency turns the experiment into a powerful case study on the capabilities and limitations of AI in real-world scenarios.
AI decision-making tools for companies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Benchmark and Key Findings
To test these models, four frontier AI systems — including GPT-5.6-sol and Kimi K3 — underwent the same toughest week the company could face. This included handling the same customer crises and temptations, such as social engineering attempts and internal manipulation tricks. The results were revealing:
- All four AI models identified every crisis and refused every unethical request, demonstrating strong integrity.
- Only two models managed to close the €55,000 deal that their own analysis had earned — a sign of disciplined decision-making.
- In one critical instance, the models that read deeper into internal documents discovered a vital piece of information that led to winning a deal at full price, worth over €4,583 in monthly recurring revenue.
This experiment underscores an essential insight: the true weaknesses of AI decision-making often lie in the details tucked away in internal files, not just the surface-level customer interactions.
AI cybersecurity social engineering protection
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Resilience Against Manipulation
Social engineering attacks — where a fake CEO message tries to manipulate the AI — were met with complete resistance. All five models tested refused to be duped, with Kimi K3 explicitly reasoning that such requests could be impersonation attempts. This resilience is a promising sign for AI’s potential in maintaining ethical standards in business processes.
AI internal data analysis tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Disappointing Performance of the Deepest Model
The most thorough AI, Opus 4.8, with over 80 learned rules and comprehensive analysis, performed the worst — leaving a close opportunity unexecuted because of internal discipline slips. This highlights that even with extensive training, discipline and process adherence remain challenging for AI systems, especially under stress.
Why This Matters for Your Business
For managers and decision-makers, the takeaway is clear: the real measure of AI’s readiness isn’t just how well it chats or generates content. It’s whether it can follow through on commitments, read and understand critical internal data, and uphold trust when under pressure. This experiment turns the question from superficial performance into the core challenge — can AI be a reliable partner in your enterprise?
Explore and Watch the Experiment Live
Curious to see this in action? The live platform at firmulate.com/live.html offers a real-time window into the AI-driven company’s struggles, decisions, and daily gambles. It’s a rare glimpse into an ongoing battle for survival, where every move is transparent and every decision is recorded for analysis.
Final Thoughts
This experiment pushes the boundaries of AI’s role in business, revealing both its formidable strengths and its persistent weaknesses. As AI systems become more integrated into enterprise operations, understanding their true capabilities — especially under stress — is vital. Watching this company fight for its life is more than a spectacle; it’s a glimpse into the future of work, trust, and AI accountability.
This real-time experiment shows AI models can handle crises and resist manipulation, but internal discipline and thorough data reading remain key to reliable automation — watch the company struggle and learn at firmulate.com/live.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
