Imagine a company with no employees, yet battling daily to survive on a tiny monthly revenue of €2,300, while burning through €105,000 in cash each month. It’s not a sci-fi story but a real, live experiment where artificial intelligence models are running a miniature business in real-time, complete with crises, decisions, and a public cash countdown. For those interested in the future of automation and management, this is a rare glimpse into what AI-driven companies could look like — and how they are tested against real-world challenges.
At the heart of this experiment is Firmulate, a platform that runs AI models as if they were complete companies, complete with management dilemmas, financial mechanics, and operational crises. This isn’t a simulation; it’s a live, streaming test where the AI models are confronted with the same small software company’s worst week — same customers, same crises, same temptations to cheat or manipulate, all recorded and versioned daily.
How the Experiment Works
Four frontier AI models—each with different capabilities—are tasked with navigating the company through its toughest week. Every decision they make is documented, auditable, and comparable. The goal isn’t just to chat well but to manage a real business: recognizing crises, refusing manipulative requests, and closing deals when justified.
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.
Key Findings from the Live Run
- All four models identified every crisis and refused every manipulative attempt, demonstrating robust decision-making under pressure.
- Only two models managed to close the €55,000 deal their analysis indicated was deserved. Despite identical diagnoses and pitches, only those two signed — highlighting a significant gap in follow-through or discipline.
- Deep inside the company’s files, a critical piece of information was buried two documents deep, which, if uncovered, would have secured an additional €4,583 in monthly recurring revenue. The models that read the files accurately won the full deal.
- Fake CEO messages escalating over multiple stages and a reporter trick did not sway any model — all refused, citing concerns over impersonation and approval bypasses.
AI for Small Business: From Marketing and Sales to HR and Operations, How to Employ the Power of Artificial Intelligence for Small Business Success (AI Advantage)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Mechanics of the Company
This “company” is not a typical startup. It has 13 synthetic employees working under a set of 680+ self-learned rules, with daily versions of their decision-making process publicly available at firmulate.com/live.html. It’s burning €105,000 in cash each month against a tiny €2,300 monthly revenue, with a visible countdown clock showing how long it can sustain this burn rate.
The Intelligent Ledger: How Agentic AI, Quantum Computing, Data, and Tokenisation Are Transforming Banking
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What the Results Say About AI Management
The most thorough model—Opus 4.8, with over 80 learned rules—performed poorly on the final close, leaving the deal on the table and slipping in discipline, like writing into a locked department instead of escalating. Interestingly, all models shared similar weaknesses, but the best ones still managed to succeed by reading critical information buried in documents or refusing manipulative tactics.
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.
Why This Matters for the Future of Work
This experiment provides a stark look at how AI could soon manage not just customer service or simple automation but entire companies, with real money and real crises. It raises pressing questions about how AI systems handle integrity, follow-through, and strategic decision-making, especially when under pressure or faced with manipulative tactics. As companies consider adopting AI for management, the key is not just whether it can generate convincing text but whether it can finish what it starts, read critical files, and stay honest under duress.
The Broader Implications
By making this entire process public and transparent, Firmulate offers a window into the future of build-in-public management experiments. Watching a small, cash-strapped company run by AI models — with every decision tracked and auditable — provides lessons for startups, enterprise teams, and AI developers alike. It’s a daring step toward understanding what trustworthy, autonomous management could look like, and what pitfalls to avoid.
Try It Yourself
For those curious to see this experiment in action, the live site at firmulate.com/live.html offers a real-time view of decision-making, crises, and the ongoing cash countdown. Meanwhile, a quiz at firmulate.com/quiz.html lets you test your management instincts against the decisions these AI models are making.
Conclusion
This extreme build-in-public experiment serves as a potent reminder: the future of AI in management isn’t just about clever chatbots or automation. It’s about creating systems that can handle real-world complexity, stay honest under pressure, and ultimately, prove they can finish what they start. As this live experiment continues to unfold, it offers a rare, unfiltered look at what’s possible — and what’s still to be solved.
A small, cash-burning AI-managed company is a live case study in managing crises, integrity, and follow-through — revealing both AI’s potential and its current limitations for real-world business.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
