Inside a Zero-Employee Software Company Fighting for Survival — Live and Unfiltered

Imagine a business that operates without a single employee, yet faces daily financial struggles while the world watches. This isn’t fiction — it’s the real-time story of Firmulate, a public experiment in AI-driven management. For fashion and style followers accustomed to every detail being curated and perfect, this story challenges notions of control, trust, and the future of work.

The Raw Reality of a No-Employee Company

Firmulate is not your typical tech startup. It’s a fully automated, publicly visible company that runs on 13 synthetic employees powered by advanced AI models. The company burns through €105,000 each month, yet earns just €2,300 in monthly recurring revenue. Every workday, its operations are versioned and publicly available, offering a rare window into how AI manages a business under pressure.

Running the Same Crisis, Different Models

Four of the latest frontier AI models — including GPT-5.6-sol and Kimi K3 — were tasked with guiding this virtual company through its worst week. The scenario was repeated identically across all models: same customers, same crises, same temptations to manipulate or cut corners. The goal? To see if these AI agents could not only diagnose and respond but also resist deception and misconduct.

What the Results Reveal

• All four models identified every crisis and refused manipulation attempts, demonstrating a remarkable capacity for integrity under pressure.
• However, only two managed to close a deal worth €55,000, which their own analysis indicated was fully earned — the others failed to follow through despite correct diagnosis.
• Interestingly, the decisive advantage came from a hidden detail within the company’s own files, not the customer interactions. The models that read this ‘buried fact’ secured the deal at full price, adding over €4,500 monthly recurring revenue.

Testing Trust and Manipulation Resistance

Beyond financials, the experiment included complex social engineering scenarios. Fake CEO messages and a reporter trick — designed to bypass trust — were issued in staged escalation. All five models refused to be duped, with Kimi K3 explicitly treating such requests as potential impersonation or approval-bypass attempts.

The Live Experiment in Action

Firmulate’s setup is both a lab and a battleground. Its ’employees’ are AI models operating in a highly detailed, version-controlled environment, where every decision is recorded and scrutinized publicly at firmulate.com/live.html. The company’s financials are real: it is burning €105,000 monthly against a tiny revenue stream, with a public cash countdown adding a sense of urgency and transparency.

Depth of Analysis and Discipline

Among the models tested, Opus 4.8 was the most thorough, analyzing over 80 learned rules and providing the deepest insights. Yet, even with this detailed approach, discipline slipped, and the close was left intentionally unexecuted. All models showed the same fundamental weakness: a tendency to avoid escalation and instead write decisions into a locked department, risking inaction during critical moments.

Implications for Business and AI

This experiment offers a stark lesson: the quality of AI decision-making isn’t measured by how well it chats or responds, but by its ability to act honestly, read relevant internal data, and follow through on commitments. For any enterprise considering AI tools—whether in sales, customer support, or planning—the question changes from “can it write well” to “will it finish what it starts, and will it stay trustworthy when under pressure?”

Why This Matters to You

Fashion and style brands, which rely heavily on image, trust, and consistent delivery, might find the lessons here surprising. AI that manages operations silently in the background must be honest, reliable, and disciplined—traits that aren’t always visible in polished demos. The transparency of the Firmulate experiment offers a front-row seat to the real risks and potential of AI-driven management.

Final Reflection

This ongoing live experiment is more than a tech spectacle; it’s a glimpse into the future of autonomous business processes. As AI models become more embedded in daily operations, understanding their strengths and weaknesses—like their ability to resist deception or follow through—will be crucial. The question isn’t just whether they can generate convincing text, but whether they can truly run a business with integrity under pressure.

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

Powered by Thorsten Meyer AI