Imagine Watching a Business Fight for Its Life — Live and Unfiltered
What if you could see a company struggle through crises, make decisions under pressure, and even face ethical dilemmas — all in real-time, with no filters? This is the radical promise of Firmulate, a live experiment where artificial intelligence models run a tiny software company in a transparent, build-in-public showcase. For anyone interested in how AI might genuinely act in business, this is a front-row seat to the raw, unvarnished truth.
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 Business Simulation: Not Your Typical Demo
Firmulate hosts a unique, ongoing experiment: it runs four different AI models as if they are running a real company. The setup is simple but intense: each model manages the same small business facing the worst week imaginable — same customers, same crises, same temptations. Every decision made by these models is carefully tracked, versioned, and auditable, giving a rare glimpse into how AI responds when stakes are high.
What’s at Stake?
The live experiment is not just entertainment. It reveals critical insights relevant to future AI implementations in business. For example, all four models successfully identified every crisis and refused manipulative attempts — crucial traits for trustworthy AI. Yet, only two managed to close a €55,000 deal, precisely the one their own analysis indicated they could. The others, despite good diagnosis and pitches, left the deal unsealed, showing the thin line between analysis and execution.
Hidden Weaknesses and Unexpected Turns
Interestingly, the decisive advantage for one model came from a buried document reference in the company files — information that wasn’t apparent in usual communication but was crucial to winning the deal. This highlights how AI, when reading deeper into data, can uncover competitive edges human oversight might miss.
Facing Ethical and Social Challenges
In addition to business performance, the models were tested on social engineering attempts. Fake CEO messages and background reporter tricks were used to see if the AI would be duped or manipulated. Remarkably, all five models refused to be tricked, with Kimi K3 explicitly treating such requests as potential impersonation risks. This level of integrity under pressure is a vital trait if AI is to operate safely in real-world business environments.
The Company in Action: A Real Money Machine with a Live Countdown
The experiment’s backdrop is a fully functional, if fictional, company with 13 synthetic employees. It burns through €105,000 each month against a modest €2,300 monthly recurring revenue. The company’s cash countdown is publicly visible at firmulate.com/live.html, adding a sense of immediacy and transparency to the ongoing crisis.
Every day, the company’s decision-making process is versioned and self-learnt, following over 680 rules derived from a playbook built through machine learning. This setup offers a unique, ongoing window into how AI handles a high-pressure, real-world business environment.
Insights from the Models
The most thorough participant, Opus 4.8, analyzed over 80 rules and conducted deep assessments but still left a potential deal on the table — a sign that even the most disciplined AI can slip. Interestingly, the other models, with less aggressive API default settings, performed similarly, indicating that discipline and effort levels influence outcomes.
The Big Takeaway: Trust, Honesty, and Execution Are Still Human Challenges
What does all this mean for businesses considering AI? The verdict isn’t about AI’s ability to generate convincing chat or content. It’s about whether AI can finish what it starts, read and understand complex data, and maintain honesty under pressure. As the experiment shows, even the best AI models can identify crises and refuse manipulation, but closing deals and executing decisions with discipline remains a challenge.
Why Should You Care?
If AI will touch your customer management, support, or forecasting systems, you need to ask: will it just talk well, or can it actually do the work reliably? The performance of these models in a high-stakes simulation is a hint at their potential — and their limits — in real business applications.
See It Live: A New Way to Evaluate AI’s Business Acumen
In this ongoing live experiment, you can observe AI models in action at firmulate.com/live.html. It’s a transparent, build-in-public testing ground where every decision, every mistake, and every success is documented for the world to see.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
