Watch a Business Live: Can AI Save a Struggling Company?
Imagine a real company operating every single workday, losing thousands of euros while its AI models are tested against crises, ethical dilemmas, and tough negotiations — all in public view. This is not science fiction, but the extraordinary experiment happening now at Firmulate, where every decision, every risk, and every outcome is open for observation.
AI-Powered Business Intelligence: Improving Forecasts and Decision Making with Machine Learning
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Setting: An AI-Driven Company on the Brink
Firmulate runs a small, real software company with a twist: it has no human employees. Instead, it employs 13 synthetic team members driven by advanced AI models. Every day, the company faces typical business crises — customer issues, sales negotiations, ethical tests — all with the added pressure of real money mechanics. Currently, it burns through €105,000 each month against a modest €2,300 monthly recurring revenue (MRR). The public cash countdown and daily versioning of decisions make this a transparent, real-time look at AI’s potential and limitations in business.
AI-Powered Cybersecurity: AI Tools for Enterprise Security | AI for Network Security | AI Risk Management | AI in Cyber Policies | Cyber Threat Management AI | ML in Fraud Prevention
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Experiment: Four Frontier AI Models in Action
Four leading AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8—were tasked with navigating the same challenging week in this simulated company. Each faced identical crises, from customer mishaps to ethical dilemmas, with decisions carefully tracked and auditable. The goal? To see how well AI could manage real-world business pressures without human intervention.
The Scalpel and the Algorithm: Reclaiming Ethical Clarity and Clinical Confidence in the Age of AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Successes, Failures, and Surprising Insights
All models successfully identified every crisis and refused to be manipulated. When it came to closing a critical €55,000 deal, only two models actually signed the contract, despite all making the same diagnosis and pitches. The missing piece was buried deep in the company’s internal files, not immediately visible in customer interactions. Those models that read and analyze this hidden information secured the deal at full price—adding over €4,500 to monthly recurring revenue.
In a separate social engineering test, the models faced fake CEO messages and a reporter’s covert request. All five models refused to escalate or approve anything suspicious, demonstrating a strong capacity for security and ethical judgment. Kimi K3, in particular, articulated a cautious approach: “Treat the request as a suspected approval-bypass / possible impersonation.”
The Project Management AI Handbook: Leveraging Generative Tools in Waterfall and Agile Environments
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Cost of Running an AI-Led Company
Despite the promising performances, the company remains in dire straits financially. It is actively losing €105,000 each month, with a small MRR of €2,300. This stark reality underscores that AI decision-making quality is only part of the puzzle—sustainability and profitability are equally crucial. The live site at firmulate.com/live.html offers a transparent window into this ongoing struggle, with every decision, crisis, and outcome publicly recorded and accessible.
Lessons from the Deep Analyses
Among the models, Opus 4.8 demonstrated the most thorough analysis, employing over 80 learned rules and detailed reasoning. However, even its deepest insights could not prevent a slip — it left a crucial closing decision unexecuted, illustrating that discipline and process adherence remain challenging for AI systems, especially under pressure.
Implications for Business and AI Adoption
This experiment shows that AI can potentially identify crises and refuse unethical requests in real business scenarios. But it also highlights significant gaps: AI models still struggle with complex decision-making, especially when it involves reading deeply hidden information or maintaining consistent discipline amid chaos.
For business leaders and investors, the key takeaway is not whether AI can write convincingly but whether it can reliably finish what it starts, read your files thoroughly, and stay honest under pressure—traits critical for real-world applications. The ongoing live experiment at firmulate.com/live.html offers a real-time demonstration of these capabilities and limitations.
What This Means for Investors and Business Owners
Watching a real company operated solely by AI models struggle, succeed, and sometimes slip, provides invaluable insights into AI’s readiness for enterprise tasks. The essential questions are: Will AI help your business finish what it begins? Will it remain honest under pressure? And, crucially, what is the true cost of AI-driven productivity in complex, profit-driven environments?
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
