Can AI Run a Company Without Employees — and Survive? Watch the Live Experiment

Imagine a business where no humans work, yet it’s constantly battling crises, making tough decisions, and losing money every day. For creators and tech enthusiasts, this isn’t science fiction — it’s the real-time experiment from Firmulate, a company run entirely by AI models. What happens when AI manages a business in its most extreme form? The answer might reshape how we think about automation, trust, and the future of work.

The Live Experiment in Action

Firmulate is not a typical startup. It operates with 13 synthetic employees, each powered by advanced AI models, and is publicly streaming its daily struggles at firmulate.com/live.html. Every workday, the company faces real crises—customer issues, internal dilemmas, and ethical temptations—just like any business. But instead of humans, decision-making falls to AI models, which are rigorously tested and versioned after each day.

The Models Behind the Decision-Making

• Four frontier AI models were given the same worst-week scenario: the same customers, crises, and temptations.
• Each model’s decisions are auditable, and they are evaluated on their ability to spot crises, resist manipulation, and close deals.
• Scores range from 77 to 95, with the top model, gpt-5.6-sol, achieving the highest performance by discovering a hidden opportunity that led to a €4,583 MRR increase.

What the Models Achieved

All four AI models successfully identified every crisis and refused every manipulation attempt, including social engineering tricks such as fake CEO messages. Yet, only two models managed to close the €55,000 deal, which their own analysis had earned, highlighting a gap between diagnosis and action.

The Hidden Weakness

One of the most revealing findings was that the decisive opportunity was buried two document references deep within the company’s internal files, not in the immediate customer interactions. The models that read and understood these files won the deal at full price, demonstrating the importance of thorough information processing.

The Company in Practice

Beyond the AI models, the ‘company’ itself is a real, operating entity with actual money mechanics. It burns €105,000 each month against a current monthly recurring revenue (MRR) of €2,300. Its funds are publicly counted down, and every decision is transparent and versioned, offering a rare glimpse into the daily grind of a business run by AI.

The Deep Dive: Opus 4.8

Among the models, Opus 4.8 stood out for its thoroughness, analyzing over 80 learned rules and producing deep insights. However, it left opportunities unexploited, such as failing to escalate issues properly, which cost it in closing deals. This shows that even the most detailed AI models can slip when discipline wanes under pressure.

Implications for the Future of Work

This experiment isn’t just about watching an AI-managed business; it raises crucial questions for anyone relying on automation. When AI agents interact with customer data, support queues, or sales pipelines, it’s not just about their ability to generate text or insights. The key is whether they can follow through, stay honest, and complete work under real-world pressures.

For creators and tech enthusiasts, this transparent, day-by-day unfolding scenario offers a glimpse into a future where AI might independently run parts of a business or creative enterprise, but only if it can navigate crises, resist manipulation, and close deals reliably.

Why You Should Care

As automation becomes more integrated into creative tech, understanding what AI can actually accomplish — and where it still falters — is vital. Will AI support your workflow with integrity and discipline? Or will it leave opportunities on the table? The Firmulate live experiment makes that question painfully clear, in real time, every day.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html

Powered by Thorsten Meyer AI