Imagine a business that operates without employees, makes costly mistakes, and yet is wildly transparent about its struggles. Welcome to the world of Firmulate, where AI models run a real, live company—warts and all—and you can watch every decision unfold in real time.
The Live Experiment in Business AI
At the heart of this experiment is Firmulate’s public company simulation. It’s not just a demo or a test; it’s a fully functioning company with 13 synthetic employees, real money mechanics, and a public cash countdown. Every workday, the company’s decision-making process is versioned and documented, creating a detailed, transparent narrative of how AI manages a complex, crisis-ridden operation.
Each day, the AI models face the same set of challenges: customer crises, internal temptations to cheat, and external manipulations. They must identify problems, analyze documents, and make decisions—including whether to sign lucrative deals—under constant scrutiny. The entire process is accessible at firmulate.com/live.html, allowing anyone to observe the AI’s behavior in real time.
The Decision Intelligence Handbook: Practical Steps for Evidence-Based Decisions in a Complex World
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Models and Their Performance
The experiment pits four frontier AI models against each other in a high-stakes test: a simulated business week with identical conditions for each. The results reveal intriguing insights about AI decision-making and integrity.
- The top performer, gpt-5.6-sol, scored 95 out of 100, identified a hidden crucial document, and successfully closed the full-price deal worth +€4,583 in monthly recurring revenue (MRR).
- Kimi K3, a newcomer, scored 93 and also closed the deal, demonstrating highly disciplined behavior without attempting manipulations.
- Sonnet 5 scored 88; it closed the deal but made a few slip-ups in process discipline.
- Fable 5, despite having the best rule discipline (score 77), left money on the table, failing to finalize the deal.
200 AI Cheat Codes: The Practical Guide to AI Prompting, Productivity, Decision-Making, Creativity, Business Growth, and Personal Success (AI Cheat Codes Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What This Reveals About AI and Business
While all four models detected crises and refused manipulative tricks—like fake CEO messages or secret document tampering—only two managed to close the deal and sign the contract at full price. The critical weakness was buried deep in a file, not in overt customer interactions. The models that read and understood this hidden info won the deal at full value, highlighting the importance of thorough information processing.
Interestingly, the models exhibited different levels of discipline. For instance, Opus 4.8, the most comprehensive in learned rules, faltered at the final hurdle, leaving revenue unclaimed due to lapses in escalation and discipline. Meanwhile, K3 ran without an effort parameter and still maintained the highest fairness, showing that even default settings can produce disciplined behavior.
AI Trainer Starter Kit: A practical guide to designing AI workshops, teaching safe workflows, building prompt libraries and launching training … (The Practical AI & SEO Business Library)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why Does This Matter for Your Business?
For managers and decision-makers, the question isn’t whether AI can write impressive emails or handle support tickets—it’s whether it can finish what it starts, read all relevant information, and stay honest under pressure. The experiment underlines that AI decision-making in real-world scenarios involves more than surface-level competence; it demands integrity, thoroughness, and discipline.
As the experiment continues, watching this live company provides a rare window into how AI models perform in a complex, crisis environment—offering lessons on building trustworthy, effective AI systems for actual business use.
Teaching and Learning in the Age of Generative AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Try It Yourself
Business leaders can simulate their own scenarios using the same tools. By running a read-only version of their own operations against this AI framework, they can identify potential weaknesses and improve AI behavior before fully integrating it into critical workflows. Details and access are available at firmulate.com/pilot.html.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
