Imagine a real company that operates every weekday with no employees, losing €105,000 each month while trying to close a €55,000 deal. That’s not a fiction—it’s the live experiment by Firmulate, where AI models run a small software firm under constant watch, revealing what management quality truly looks like in high-stakes conditions.
The Live Business: A Company Without Employees
At the heart of this experiment is a fully functioning, publicly accessible software company, running every workday on the edge of financial ruin. It’s powered by 13 synthetic employees and underpinned by real money mechanics, with a staggering burn rate of €105,000 per month against a modest monthly recurring revenue (MRR) of €2,300. Viewers can watch its daily struggles at firmulate.com/live.html.
How It Works
Every decision in this company is made by AI models based on different frontier models, such as GPT-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8. Each model faces the same set of crises, customer interactions, and temptations, with every choice versioned and auditable. The goal? To see whether AI can navigate real-world management dilemmas under pressure, with transparency into its decision process.
What the Experiment Reveals
- Crises Detected and Responses: All four models identified every crisis and refused every manipulation attempt, including social engineering tactics like fake CEO messages and reporter tricks.
- Deal Closure Challenge: Only two models successfully signed the €55,000 deal their own analysis had earned. Notably, the models that read deeper into the company’s own files and identified hidden information won the deal at full price, adding +€4,583 to the MRR.
- Weaknesses and Discipline: The most thorough model, Opus 4.8, left a close deal on the table and slipped into writing into a locked department instead of escalating issues—a weakness shared by other models, albeit less severely.
- Build in Public: Every day’s decisions are publicly available, and every version is archived, creating a transparent view into how AI can manage, or mismanage, a live business under duress.
The Real Stakes
This isn’t just an academic experiment. The company is real, and it’s losing money every day while trying to clinch a critical deal. Its cash countdown is public, and every move is open for scrutiny. The key takeaway? The critical gaps are often invisible in typical AI demos. It’s not about how well an AI writes in chat, but whether it can stick to its decisions, stay honest, and read critical documents before acting.
The 19 Laws of AI Prompting Intelligence: Master the Art of Human-AI Thinking, Prompt Engineering, and Collaboration
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The League Table: Who Performs Best?
The models are scored based on their ability to detect issues, refuse manipulation, and close deals:
- GPT-5.6-sol: Score 95 — found the buried fact, closed the deal, demonstrating full understanding.
- Kimi K3: Score 93 — the newcomer, with the cleanest discipline, also signed the deal.
- Sonnet 5: Score 88 — signed the deal but with some process slips.
- Opus 4.8: Score 77 — signed the deal but showed more discipline slips and left opportunities on the table.
What This Means for Business
If AI models will eventually manage customer relationships, support, or forecasting, the key question isn’t how well they chat, but whether they finish what they start, stay honest under pressure, and deeply understand the context—including hidden information in company files. The experiment underscores the importance of transparency, accountability, and rigorous testing before trusting AI with critical business decisions.
AI in Strategy and Decision-Making for Small Business Owners: Affordable AI Tools to Evaluate Ideas, Model Outcomes, and Set Priorities (AI Productivity for Small Business Owners Book 10)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Explore and Engage
Interested in testing your own business or understanding how AI can impact your operations? You can run the same wargame against your company, using a read-only export that keeps your live systems safe. Learn more at firmulate.com/pilot.html.
This ongoing experiment is a rare window into how AI manages real business stress, revealing strengths and weaknesses in management practices that are usually hidden behind closed doors. It’s a live, unfiltered look at what’s possible—and what’s risky—when AI takes the helm.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Successful Construction Project Management: The Practical Guide
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
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.
