Imagine a business where every decision is made by AI, yet it operates in the harshest conditions — facing crises, temptations, and the relentless pressure of survival. This isn’t fiction; it’s the groundbreaking live experiment from Firmulate, where a simulated company is fighting for its life in full public view, revealing the true capabilities and limitations of AI decision-making in real-world scenarios.
The Live Experiment: An AI Company Under Pressure
At the heart of this bold venture is a small, synthetic business with 13 AI-driven employees, tasked with running a software company through its worst week. Every decision, crisis, and temptation is transparently logged, versioned daily, and accessible to the public. The company’s mechanics are real: burning €105,000 monthly against a modest €2,300 in monthly recurring revenue (MRR), with a public countdown ticking down its cash supply.
What makes this experiment extraordinary is not just its transparency but the rigor of its setup. Four leading AI models — GPT-5.6, Kimi K3, Sonnet 5, and Fable 5 — were all subjected to identical scenarios: same customers, same crises, same opportunities to cheat. Their decisions are fully auditable, and their performance measured objectively.
AI decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: AI’s Capabilities and Limits
- All four models successfully identified every crisis, demonstrating a strong grasp of immediate problems.
- None of the models succumbed to manipulation attempts, including social engineering tactics like fake CEO messages and reporter tricks, which all were refused — notably, Kimi K3 explicitly flagged impersonation risks.
- The critical weakness was in the details buried two document references deep within the company’s own files. Reading these internal documents enabled the models to identify a hidden opportunity: a deal worth over €4,583 in additional monthly revenue, which three models missed but two seized — closing the deal at full price.
- Only two models signed the €55,000 deal that their own analyses confirmed they should have. The other two, despite correct diagnosis and pitch, left the deal unclosed, illustrating discipline lapses and process slip-ups.
business crisis management tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Real Money and Real Risks — Live in Public
The experiment isn’t just a test; it’s a public showcase of an AI company on the brink, with transparent financials and every decision scrutinized. The company burns through €105,000 a month and has just €2,300 in MRR, with a visible cash countdown — every day a reminder of its fragile survival.
The platform allows viewers to watch this ongoing struggle at firmulate.com/live, where every workday is versioned and documented, offering a rare glimpse into AI decision-making under true business pressures.
AI risk assessment tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Implications for the Future of AI in Business
This experiment underscores a crucial point for organizations deploying AI: passing demos isn’t enough. It’s about whether AI can consistently finish what it starts, read and interpret internal documents, stay honest under pressure, and make decisions that truly add value.
The models showed remarkable competence in crisis detection and integrity — but also revealed process weaknesses in closing opportunities, especially when discipline slipped or rules were not fully applied. The real-world takeaway is clear: AI can be a formidable business partner, but only if it’s tested rigorously against real pressures and temptations.
internal document analysis software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Broader Message: Build in Public, Build with Rigor
Firmulate’s live experiment exemplifies the extreme end of transparency and continuous testing — a ‘build-in-public’ approach taken to the utmost. It proves that AI’s true readiness for business isn’t just about generating content or insights but about executing complex, high-stakes decision workflows reliably and honestly.
For leaders and technologists alike, the message is straightforward: before trusting AI to run your operations, put it through its paces in real-time, public scenarios where every decision is watched, analyzed, and learned from.
Watching a live AI-driven company struggle and succeed in real time reveals crucial insights: AI can identify crises and resist manipulation, but discipline in closing deals and process adherence remains vital. This public experiment highlights the future of trustworthy AI in business — tested, transparent, and relentlessly pushed to perform.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
