What if you could see AI in action managing a real company, facing crises, making decisions, and even losing money — in real time?
Imagine a world where you can watch artificial intelligence models run a small software business, caught in the chaos of daily crises, battling to keep afloat, and yet, sometimes, missing the chance to close a deal. This is no fiction — it’s the live experiment from Firmulate, a groundbreaking project that puts AI management to the test in front of a live audience.
AI Builders: Making The Decisions That Turn AI Code Into Real Software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Live Company in Action
At the heart of this experiment is a simulated, yet fully operational, company with 13 synthetic employees, each governed by an ever-evolving set of 680+ self-learned rules. This company is not just a digital sandbox — it’s a real, money-burning enterprise, losing €105,000 every month against a modest €2,300 monthly recurring revenue (MRR). The public can watch every decision, every crisis, every rule applied, and every version of the company as it progresses — every workday is versioned, archived, and transparent.
The Experimental Setup
Four state-of-the-art AI models, representing different frontiers of language and decision-making technology, were each tasked with running this struggling company through its worst week — the kind of week filled with customer crises, tempting manipulations, and complex negotiations. Each model was given the same set of challenges: the same customers, the same crises, and the same opportunities for manipulation. All decisions were fully auditable, and the models were tested against real-world tactics such as social engineering and document analysis.
Surprising Results
While every model successfully identified and responded to crises — a crucial first step — their ability to close deals varied dramatically. The most successful, GPT-5.6-sol, scored 95 out of 100, identified the buried opportunity hidden two document references deep in the company’s files, and closed the deal worth €55,000, adding +€4,583 to monthly recurring revenue. The second, Kimi K3, scored 93 and also secured the deal, demonstrating the discipline and integrity that makes it the standout in the field. The others, despite recognizing crises, failed to follow through and left significant opportunities unexploited.
What Does This Mean for AI in Business?
The experiment underscores a critical point: AI’s true test in business isn’t just generating convincing chat or reports. It’s about executing real work — reading crucial documents, making decisions under pressure, resisting manipulation, and ultimately, closing deals. The models proved they could diagnose crises and refuse social engineering attempts, a vital aspect of trustworthy AI. Yet, without proper discipline and a focus on execution, even the most capable AI can falter.
The Live Experiment — A Window into the Future
Firmulate’s live site offers a rare glimpse into this world. You can observe the AI-driven company in action, watching as decisions unfold, crises are tackled, and deals are either made or missed. Every day offers new insights into what AI can and cannot do as a management tool. It’s a fascinating, transparent look at an ongoing struggle for survival in the age of automation, where the question isn’t just about AI’s intelligence but its integrity and discipline.
Implications for Business and AI Deployment
For businesses cautiously exploring AI tools, this experiment is a stark reminder: the difference between a helpful AI and a risky one depends on how well it can see the full picture, stay disciplined, and follow through. AI models can spot crises and refuse manipulation, but closing deals and executing plans require a level of thoroughness and discipline that is still a challenge. The question is not whether AI can write well — it’s whether it can finish what it starts, read critical documents, and stay honest under pressure.
Explore and Test for Yourself
Interested in how AI might perform in your own business? Firmulate offers a unique opportunity: run a wargame against your company’s data, test AI decision-making, and identify weaknesses before actual deployment. This transparent, no-write-back approach allows you to see how AI handles crises and opportunities without risking your real systems.
The experiment vividly demonstrates that AI management is more than just smart responses. It’s about execution, integrity, and discipline. Watching AI companies in crisis — with real money and real risks — reveals both their promise and their current limits.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
