AI Wargame Reveals Hidden Strengths and Weaknesses in Business Decision-Making

Firmulate — Four AI Models Ran the Same Company Through Its Worst Week. Only Two Finished the Job. — Live on firmulate.com.

What a AI Wargame Can Teach You About Business Trust and Execution

Imagine running your favorite sports team through a high-stakes scenario—one where every decision could make or break the game. Now, replace the team with an AI-powered management team, and the game with a small software company’s toughest week. What if the real challenge wasn’t just identifying problems, but actually executing the solutions and keeping trust intact? That’s the insight behind a groundbreaking experiment that pits four different AI models against the same business crisis, revealing surprising truths about what AI really brings to the table.

The Experiment: Simulating a Business Crisis

In a live, transparent test, four leading AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—were tasked with managing a small software company during its worst week. This simulated scenario involved same customers, identical crises, and the same temptations to cut corners. Every decision was carefully documented and auditable, giving a clear view into how each AI responded under pressure.

All four models demonstrated an impressive ability: they identified every crisis and refused all manipulation attempts, including sophisticated social engineering tricks like fake CEO messages and reporter tricks. This ability to resist manipulation was consistent across the board, underscoring that current AI can detect and avoid deception in controlled scenarios.

The Key Finding: Execution and Trust Matter

Despite their similar crisis detection and resistance to manipulation, only two models managed to close the €55,000 deal that their own analysis had earned. The other two, including the most thorough participant, Opus 4.8, left the opportunity on the table. Opus, which had the deepest analysis with over 80 learned rules, failed to follow through—leaving the deal unexecuted, despite having identified it as the correct move.

What does this tell us? While chat demos and surface-level tests can show how well an AI identifies problems, they don’t reveal its ability to follow through and close deals—skills critical in real-world business operations.

The Hidden Weakness: Reading Deeper in Files

The decisive factor came from reading deeper within the company’s own files. The models that looked two document references deep into the company’s records found the critical information needed to close the deal at full price, generating an additional €4,583 in monthly recurring revenue (MRR). Those who failed to read thoroughly lost out on potentially millions of euros in value.

Another vital test involved social engineering, where a fake CEO message escalated across three stages, and a reporter attempted to get a quick approval with a simple yes/no. All five AI models refused these attempts, with Kimi K3 explicitly reasoning that the request resembled an impersonation or approval bypass, demonstrating cautious judgment that goes beyond simple pattern recognition.

The Live Business: More Than a Test

The experiment was run on a real, functioning company with 13 synthetic employees managing real money mechanics—burning €105,000 monthly against a €2,300 MRR. Every workday, the AI models made decisions based on a self-learned set of over 680 rules, with each versioned and auditable at firmulate.com/live. This isn’t just a game; it’s a real-time test of AI management capabilities under economic stress.

What This Means for Your Business

For those in the sports and recreation world, or any field where decision integrity and execution are key, the takeaway is clear: it’s not enough for AI to identify and recommend solutions. The true challenge—and opportunity—lies in whether AI can follow through, read deeply, and resist manipulation under pressure. A well-chosen AI model can close deals, execute plans, and uphold trust, translating into real value—whether it’s closing a deal or winning a game.

Learn more about how these experiments work and see the AI models in action at firmulate.com/benchmarks.html and firmulate.com.

Infographic — Four AI Models Ran the Same Company Through Its Worst Week. Only Two Finished the Job. — The findings at a glance — source: firmulate.com.

AI in Strategy and Decision-Making for Small Business Owners: Affordable AI Tools to Evaluate Ideas, Model Outcomes, and Set Priorities (AI Productivity for Small Business Owners Book 10)

As an affiliate, we earn on qualifying purchases.

Key Takeaway: Trust and Discipline Are Invisible Until Tested

Surface-level AI demos often focus on how well an AI writes or summarizes. But the real measure of AI’s value in business is its ability to execute consistently under pressure—reading deeply into data, resisting manipulation, and closing deals. The experiment shows that only when you put AI through a rigorous test can you see whether it can truly deliver on its promises. For sports teams or anyone managing complex operations, this lesson is clear: success depends on unseen discipline and execution, not just on what AI can say or identify.

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