AIThis post was created with the assistance of artificial intelligence (AI).
Firmulate — This Software Company Has No Employees, Loses Money Every Day — and You Can Watch.
Live on firmulate.com.

Imagine a business that’s live, transparent, and running in real time — with no human employees, billions of dollars at stake, and every decision scrutinized.

For the first time, a software company is operating publicly with a simulated team of AI executives, navigating crises, ethical dilemmas, and strategic decisions every single day. This is not a futuristic fantasy; it’s happening now, offering a behind-the-scenes look at how artificial intelligence can manage a real business under extreme pressure.

The Decision Intelligence Handbook: Practical Steps for Evidence-Based Decisions in a Complex World

The Decision Intelligence Handbook: Practical Steps for Evidence-Based Decisions in a Complex World

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

The Live Experiment: An AI Company in Action

At the heart of this experiment is a fully operational online platform that runs a small software company, but with a twist — the company’s 13 employees are entirely synthetic, powered by advanced AI models. Each day, these models face the same challenging scenarios: customer crises, internal conflicts, and ethical tests. Their decisions are recorded, versioned, and publicly accessible, offering a rare window into AI decision-making in a business context.

How It Works

The company’s daily operations are run by four different frontier AI models, each with its own strengths and weaknesses. These models are subjected to identical crises — from technical failures to manipulative social engineering attempts. Remarkably, all four models identified every crisis and refused all manipulative requests, demonstrating a high level of integrity and situational awareness. Yet, only two managed to close a critical deal worth €55,000 — the same diagnosis and pitch, but divergent outcomes.

What Decides Success?

The key difference in performance came from how the models handled internal information. The winning models read deeper into their own files, uncovering a crucial document that contained the secret to closing the deal at full price, adding €4,583 to monthly recurring revenue. In contrast, the others missed this buried fact and left the opportunity on the table.

Ethics Under Pressure

The system was also tested against social engineering attacks. Fake CEO messages and a staged journalistic request attempted to bypass controls and sway decisions. Every model refused these manipulative tactics, with Kimi K3 explicitly treating suspicious requests as potential impersonation, reflecting a cautious, security-conscious approach.

The Reality of a Money-Losing Business

Despite the sophistication, the live AI company is far from profitable. It’s burning €105,000 each month against a mere €2,300 in monthly revenue, with a public countdown to its cash reserves. Every day, the business operates under a strict set of self-learned rules — over 680 of them — and every decision is carefully recorded and available for review at firmulate.com/live.html.

The Deep Dive: Model Performance and Lessons

The most thorough participant, Opus 4.8, analyzed over 80 rules and performed the deepest assessments. Yet, it still left potential deals unclosed, slipping into procedural slips rather than strategic mistakes. This highlights how even advanced AI models can stumble under real-world pressures, especially when disciplined execution falters.

Implications for Business and AI

This experiment underscores a vital point: for AI to be truly valuable in business, it must do more than generate convincing chatter. It must reliably finish what it starts, read and understand relevant internal documents, and remain honest under pressure. As AI begins to touch customer relationships, support operations, and forecasting, these qualities will determine its real utility and trustworthiness.

Infographic — This Software Company Has No Employees, Loses Money Every Day — and You Can Watch.
The findings at a glance — source: firmulate.com.

Key Takeaways

This live experiment reveals that AI can identify crises and avoid manipulation, but closing deals and maintaining discipline remain challenging. For businesses considering AI for decision-making, the focus should be on reliability, integrity, and the ability to read and act on internal data — not just generating appealing conversations. Watch this real-time struggle for survival unfold at firmulate.com/live.html, and see what the future of AI-driven companies might look like.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html

Powered by Thorsten Meyer AI


You May Also Like

Repairing Loose Pool Tiles: Tools and Techniques

Caring for loose pool tiles requires specific tools and techniques that ensure a durable fix—discover how to restore your pool’s appearance and safety.

An “Unusable and Overgrown” Yard Got a Dreamy Patio Redo with DIY Furniture

A neglected, overgrown yard was turned into a functional outdoor space with DIY furniture and professional landscaping, enhancing the home’s usability.

How to Stop Pool Equipment Vibrations From Shaking the Pad

AIThis post was created with the assistance of artificial intelligence (AI).To stop…

Detecting Underground Plumbing Leaks With Pressure Testing

AIThis post was created with the assistance of artificial intelligence (AI).To detect…