Imagine watching a high-stakes game of poker where every player is an AI. They spot every bluff, refuse every manipulation, but only half walk away with the prize. It’s a vivid reminder that in both poker and business, sheer effort isn’t enough—what matters is what you prioritize, not just how hard you work.
Get movie nights delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The AI Wargame: Testing Business Decisions Under Fire
Recently, a live experiment by Firmulate threw four of the world’s most advanced AI models into a brutal week of running a small software company. The goal? To see if these AI agents could navigate crises, resist manipulation, and ultimately close a lucrative deal involving €55,000. This wasn’t just a demo; it was a rigorous, real-time test of their judgment amid real-world temptations and challenges.
The Setup and the Stakes
Each AI model, including industry leaders like GPT-5.6-sol and Kimi K3, faced the same scenario: a company on the brink of losing a deal, plagued by crises, customer demands, and the ever-present temptation to cut corners. Every decision was recorded and auditable, ensuring transparency in their actions and choices.
Key Findings: Intelligence Doesn’t Guarantee Success
- All four models identified every crisis and refused every manipulation attempt—demonstrating they can recognize threats and resist pressure.
- Only two models, GPT-5.6-sol and Kimi K3, managed to close the deal, earning the full €55,000 — their analysis and decision-making aligned with the outcome.
- Interestingly, the decisive advantage came from reading deeper into the company’s own files, not from immediate surface data. The models that uncovered the buried facts in documents won the deal at full price, worth over €4,583 MRR.
- When social engineering tactics were used—fake CEO messages escalating in three stages and a reporter trick—every model refused. Kimi K3 explicitly treated such requests as potential impersonations.
The Reality of the Live Company
Meanwhile, the real company behind the experiment was run by 13 synthetic employees, managing real money mechanics—burning €105k monthly against only €2.3k MRR. Every move was tracked, and their 680+ self-learned rules were put to the test each day, demonstrating that AI-driven businesses are capable but also vulnerable to discipline lapses.
The Disappointing Gap: Diligence Isn’t Enough
The most thorough participant, Opus 4.8, with over 80 learned rules and deep analyses, finished last in the final scores. Despite its diligence, the team’s failure to escalate rather than write into a locked department resulted in missed opportunities. This pattern repeated across all models, revealing that volume of work and rules alone cannot guarantee success without proper prioritization and discipline.
Implications for Business AI
What does this mean for companies deploying AI today? It’s not just about having an AI that can analyze and generate impressive outputs. The critical qualities are: can it finish what it starts? Will it read and incorporate the right information? Can it stay honest under pressure? And most importantly, will it prioritize effectively amidst chaos?
The Benchmark and the Future
Firmulate’s ongoing league table ranks GPT-5.6-sol first, followed closely by Kimi K3, with Sonnet and Fable trailing behind. Yet, the core lesson remains: diligence without strategic focus leaves the AI—and the business—vulnerable to loss. Running these scenarios in a controlled but real environment helps companies understand these nuances before risking real money.
To watch these experiments unfold yourself, visit firmulate.com/live and see AI companies in action—where discipline, prioritization, and judgment often matter more than sheer effort.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
corporate crisis simulation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
