A Glimpse into the Future: AI as Business Manager
Imagine a company without human employees, making real decisions, facing crises, and losing money daily — and you can watch it live. This is not science fiction but a pioneering experiment in AI transparency and resilience, revealing what it takes for artificial intelligence to manage real-world business challenges.
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.
The Live Company: An Unprecedented Experiment
At the heart of this experiment is a real, operational software company run entirely by AI models. There are no human employees—only 13 synthetic ’employees’ making decisions every workday, with every move public and auditable. This company spends €105,000 each month to burn through its cash reserves, earning just €2,300 in monthly recurring revenue. It’s a fragile setup, openly confronting the harsh realities of business survival.
Every day, the company’s decision-making is versioned, and its performance is measured against a leaderboard. The goal? To see whether these AI models can handle crises, seize opportunities, and stay honest under pressure. The site, firmulate.com/live, offers a watchable window into this ongoing struggle.
The AI-Driven Leader: Harnessing AI to Make Faster, Smarter Decisions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Do the Models Fare?
The experiment pits four top AI models against the same set of challenges. Among them, GPT-5.6-sol scored the highest with a 95, successfully uncovering a critical fact buried in company files that, when acted upon, secured a €55,000 deal—adding €4,583 in monthly revenue. The other models, including Kimi K3, Sonnet 5, and Opus 4.8, also attempted to close deals but with varying success.
While all four models detected crises and refused manipulative social engineering attempts—such as fake CEO messages or secret approvals—only two actually signed the lucrative deal. The key difference? The winning models delved into the company’s own documentation, uncovering hidden opportunities that others missed.
As an affiliate, we earn on qualifying purchases.
What Does This Say About AI in Business?
This experiment underscores a crucial point: AI’s ability to read and understand internal documents can be decisive. The models that examined the company’s internal files before acting were more successful, demonstrating that effective AI management isn’t just about surface-level responses but about deep understanding and integrity.
For example, during a staged social engineering attack, all models refused to approve unauthorized requests. Kimi K3, in particular, highlighted the importance of suspicion and verification, treating such requests as potential impersonation. This focus on trust and discipline is vital in real-world applications where AI must safeguard sensitive operations.
AI Phishing, Social Engineering & Fraud: How Criminals Use AI to Manipulate, Steal & Deceive (The AI Cybersecurity)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Reality of Running an AI-Managed Business
Behind the scenes, this is a real business with real money mechanics. It ‘loses’ €105,000 every month against a tiny revenue stream, and every decision is scrutinized and versioned. The company’s ongoing struggle to stay afloat is openly documented, providing invaluable insights into the capabilities and limitations of AI decision-making in complex environments.
Notably, the most thorough participant, Opus 4.8, analyzed over 80 learned rules and conducted deep assessments but still left a potential deal on the table due to lapses in discipline—such as writing issues instead of escalating problems. This highlights that even with extensive training, AI models can falter without proper oversight and process discipline.
Implications for Business and AI Adoption
Ultimately, this experiment shifts the conversation from whether AI can generate convincing chat responses to whether it can reliably complete meaningful work under pressure. If AI agents will someday handle CRM, support, or forecasting, the key questions are: Will they finish tasks? Will they read critical internal information? Will they stay honest? And at what cost?
For companies contemplating AI integration, the answer lies in rigorous testing and transparent evaluations—like this live experiment. It’s not enough for AI to mimic human conversation; it must demonstrate integrity, thoroughness, and resilience amid real crises.
Join the Movement: Test Your Own Business
Interested enterprises can run their own wargames against a read-only export of their operations, without risking real systems. Such testing offers a safe way to evaluate how AI might perform in actual business scenarios before making commitments. Details and access are available at firmulate.com/pilot.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
