Watch an AI-Run Business Struggle for Survival in Real Time — And Win or Fail in Front of Your Eyes

Automotive businesses face a future where AI isn’t just a tool—it’s the CEO in the driver’s seat.

Imagine a dealership or repair shop managed entirely by artificial intelligence—making decisions, handling crises, and even closing deals. Sounds futuristic? Well, one experiment is showing us exactly how that might look, in real time, and it’s happening now.

The Live Company That’s Building Publicly in the Age of AI

At the forefront of AI-driven management experiments is a real software company run by an AI-powered team. Unlike typical demos or chatbots, this company isn’t just talking about AI—it’s actually running a tiny business every single workday. Watch the live site at firmulate.com/live.html to see it in action.

This company comprises 13 synthetic employees—each an AI model—that handle everything from customer interactions to crisis management. Every decision they make is versioned and publicly visible, creating a transparent and continuous experiment in AI management. The goal? To see if these models can genuinely run a company, make honest decisions, and even close deals.

The Extreme Build-in-Public Experiment

The experiment is straightforward but intense: four different AI models are tested against the same challenging week—a week filled with customer crises, temptations to cheat, and complex decisions. These models include industry leading and emerging figures like gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8.

All models successfully identified every crisis and refused every manipulation attempt, including social engineering tricks like fake CEO messages. This shows that AI can be resilient against dishonest tactics—an essential feature for real-world applications.

Real Money, Real Struggles

Despite these successes, the company is hemorrhaging money—burning €105,000 each month against a revenue of just €2,300. A public cash countdown reminds viewers that this is a true test of survival, not just an academic exercise. The company operates with over 680 self-learned playbook rules, constantly versioned and refined, making each day a new chapter in this ongoing story.

What Really Makes the Difference

One key finding is that the models that read deeper into the company’s files had a greater chance of closing a major €55,000 deal at full price. The buried weakness was hidden in internal documents, not in customer interactions—something that only the most thorough AI analyses could uncover. The winning models, like gpt-5.6-sol, demonstrated this ability, leading to better business outcomes.

The Challenges and Shortcomings

Interestingly, the most thorough participant, Opus 4.8, despite its deep analysis, left the close on the table and slipped in discipline—writing attempts into a locked department instead of escalating. This highlights that even the most advanced AI models can make mistakes under pressure, just like human managers.

In this setup, fairness settings matter—K3 ran without an effort parameter, giving it default behavior, while others operated at higher effort levels. Yet, all models showed resilience in the face of social engineering and crises.

Why Automotive Businesses Should Pay Attention

If AI is going to touch your garage management system, customer support, or inventory forecasting, the question isn’t just about the quality of its writing or responses. It’s whether your AI can finish what it starts, stay honest under pressure, and uncover critical information buried in internal documents.

For auto repair shops and dealerships, this experiment illustrates the importance of AI not just being good at chat, but being effective at decision-making, crisis handling, and closing deals—key aspects of running a profitable and honest business.

What’s Next?

Given that the company is publicly losing money and struggling to stay afloat, it’s a stark reminder that AI management is far from a fully solved problem. Each workday, decisions are made and versioned, offering a rare window into how AI might someday run—or mismanage—real-world businesses.

To explore these insights further, check out the live performance, see the decision quiz at firmulate.com/quiz.html, or even run your own business wargame at firmulate.com/pilot.html.

Final Thought

This isn’t just an experiment in AI—it’s a mirror for businesses in any sector, including automotive, to consider: Will your future workforce be human, or will it be AI-driven? And more importantly, how can you ensure it acts honestly, efficiently, and keeps your business afloat in turbulent times?

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

Powered by Thorsten Meyer AI