Inside a Zero-Employee AI Company That’s Burning Cash and Publicly Testing Its Survival

Firmulate — This Software Company Has No Employees, Loses Money Every Day — and You Can Watch. — Live on firmulate.com.

Imagine a company with no human employees, losing €105,000 every month, yet still operating under the watchful eye of the world. This isn’t fiction — it’s the real-time experiment of Firmulate, a live AI-driven business emulation that’s pushing the boundaries of automation, trust, and transparency.

The Live Experiment: Running a Company Without Humans

At the heart of this experiment is a small software firm operated entirely by artificial intelligence models, which act as 13 synthetic employees. Instead of human decision-makers, these models face the same week of crises, customer demands, and temptations that real businesses encounter. Every move is versioned and documented, providing a rare window into how AI handles complex, real-world management challenges.

What makes this setup extraordinary is its transparency. The entire process is public, hosted at firmulate.com/live.html, allowing anyone to watch the company’s daily struggles, decisions, and results unfold. With a monthly cash burn of €105,000 against a minimal revenue of €2,300, the company is literally racing against its own survival, creating a dramatic backdrop for testing AI management capabilities.

The AI Models and Their Performance

gpt-5.6-sol: Scored the highest with 95 points, it identified critical information buried deep in internal files, leading to a successful €55,000 deal, adding +€4,583 in monthly recurring revenue.
Kimi K3: The newcomer scored 93, signed the deal based on its analysis, and demonstrated the cleanest discipline among all models.
Sonnet 5: With 88 points, it also secured the deal, though with slightly more process slips.
Fable 5: Despite having the best rule discipline (77 points), it left the deal unexecuted, demonstrating how discipline alone isn’t enough without decisive follow-through.

All four models detected every crisis and refused manipulative social engineering tactics, such as staged CEO messages or reporter tricks, maintaining integrity under pressure. This indicates that these AI agents are not just good at chatting but are genuinely managing the company’s operations with honesty and precision.

The Hidden Weakness and the Critical Find

The real weakness was not immediately obvious. It was buried two document references deep within the company’s files. Only the models that read the internal files fully managed to uncover this vital piece of information, which was key to closing the full-price deal. This highlights a crucial aspect of AI management: reading and understanding internal documentation can be the difference between success and failure.

Project Management with AI For Dummies

As an affiliate, we earn on qualifying purchases.

Why This Matters for Business and Tech

For entrepreneurs and managers, the lesson is clear: the quality of an AI’s work isn’t just about how well it communicates or follows commands. It’s about its ability to finish tasks, read critical internal data, and stay honest when under pressure. As AI models become more integrated into real-world business processes—support, sales, forecasting—their capacity to deliver consistent, truthful results will determine their value and trustworthiness.

The Broader Implications

The experiment shows that AI can detect crises and reject manipulation attempts, including social engineering attacks, with perfect accuracy.
Despite losing money daily, the company’s live story offers investors and AI developers a unique view into how AI can be tested against the toughest management scenarios.
Every decision and interaction is publicly available, fostering transparency in what has traditionally been a black box process.

Amazon

internal document analysis AI

As an affiliate, we earn on qualifying purchases.

The Future of AI-Driven Business

This initiative by Firmulate is more than a novelty. It’s a glimpse into a future where AI management tools could be used to simulate, test, and improve real organizations before committing actual resources. Companies could run such wargames internally, evaluating how their AI agents handle crises, negotiations, and ethical dilemmas — all without risking their actual business.

In this extreme build-in-public experiment, the AI company is fighting for survival in plain sight, revealing both its strengths and vulnerabilities. The takeaway? As these models evolve, their ability to read, decide, and stay honest might become the ultimate measure of their potential to augment or even replace human decision-makers.