Imagine an AI managing a real company, making decisions that cost thousands of euros daily, and facing crises with zero human intervention. This is not science fiction — it’s the live experiment from Firmulate, where AI models face the brutal realities of running a business in real time.
Inside the Live Experiment: AI as a Company CEO
At the heart of this groundbreaking project is a small, synthetic company operated entirely by AI models. These models are tasked with managing everything from customer crises to financial decisions. Every move they make is recorded, versioned, and subjected to rigorous scrutiny. The experiment runs in real-time at firmulate.com/live, offering a rare glimpse into the potential and limitations of AI in high-stakes decision-making.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Challenging AI with Crisis and Temptation
The models face the same tumultuous week that would cripple any human-run business. They encounter identical customer issues, internal crises, and ethical temptations. For example, they’re tested against social engineering attempts, including fake CEO messages and reporter tricks designed to manipulate decision-makers.
The results are revealing: all four AI models identified every crisis and refused manipulation attempts. Notably, when it came to closing a significant €55,000 deal, only two managed to sign at full price — despite all of them diagnosing the same opportunity and delivering similar pitches. The critical difference? The models that read deeper into the company’s own files uncovered a key piece of information that gave them the winning edge, showing the importance of thorough information processing.
The AI Culture Blueprint: Moving Beyond Tools to Create Human-Centered AI Adoption
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Real Cost of AI Management
This experiment exposes a stark reality: running this tiny company burns €105,000 every month against a mere €2,300 in recurring revenue. The company is on a public cash countdown, illustrating the delicate balance between AI decision-making and financial sustainability. It’s a vivid reminder that AI’s role in business isn’t just about generating convincing chat responses, but about effective, honest, and financially sound management.
Trauma Doesn't Stop at the School Door: Strategies and Solutions for Educators, PreK–College
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Performance and Pitfalls of Different AI Models
The experiment pits four frontier AI models against each other, with scores ranging from a high of 95 to a low of 77. The top performer, gpt-5.6-sol, successfully uncovered critical hidden facts and closed the deal, demonstrating the potential of advanced models to handle complex, real-world business tasks. Meanwhile, other models, despite their strong rule discipline, left opportunities unexploited or failed to follow through on decisions, highlighting areas for improvement.
The AI Prompt Playbook for Excel & Financial Analysis: 50 Ready-to-Use AI Prompts for Excel, Financial Analysis, Dashboards, Power BI, Financial Modeling & Automation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Build-in-Public Approach: Transparency and Innovation
What makes this project particularly notable is its commitment to transparency. Every decision, every rule learned, and every version update is publicly accessible. The company’s own playbook of over 680 rules is continuously refined, with every day’s work documented and versioned. This approach offers a valuable resource for anyone interested in AI’s evolving capabilities and limitations in operational contexts.
Why This Matters for the Future of AI in Business
For business leaders and tech enthusiasts alike, the key takeaway is clear: the question isn’t whether AI can produce convincing chatbots but whether it can reliably complete complex, high-stakes tasks. These models are tested here under real pressure to see if they can stay honest, read important files, and finish what they start — qualities critical to deploying AI in customer service, support, or decision-making roles.
As firms consider integrating AI into their operations, experiments like this serve as a reality check. They emphasize the importance of testing AI in simulated but realistic environments before committing to live deployment. The ongoing live experiment at firmulate.com/live offers a transparent window into this challenge, providing insights into what it takes to build trustworthy AI systems capable of managing real-world business complexity.
This live experiment demonstrates that AI can identify crises and refuse manipulation, but financial sustainability and thorough decision execution remain challenges. Watching AI manage a company in real time offers critical lessons for the future of trustworthy, autonomous business management.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
