Imagine preparing for a big sports event. You’ve got your gear, your team, and last-minute game plans. But what if the real deciding factor wasn’t your equipment, but whether an AI could read and understand the hidden details in your game tape — buried two layers deep? In the world of business, the stakes are just as high. AI that can dig into complex files before making decisions isn’t just a fancy feature; it’s a game-changer.
Cracking the Code: The Hidden Depths of AI Decision-Making
In a recent experiment conducted by Firmulate, four advanced AI models were put through a simulated week of crises faced by a small software business. The goal? See if they could identify critical hidden information buried two references deep in company files — and whether that knowledge would influence their decisions.
What Made This Test Different?
- Same set of crises, same customers, same temptations, for all models.
- Decisions were fully versioned and auditable, ensuring transparency.
- Models had to navigate complex scenarios, including social engineering attacks and manipulations.
The Surprising Power of Deep Reading
All four models successfully spotted every crisis and refused manipulation attempts, showing they could recognize threats and stay honest under pressure. But here’s the catch: only two models closed the deal and signed a €55,000 contract — the same amount as their own analysis indicated they should. The other two missed the buried fact, leaving money on the table.
The Hidden Factor: Deep in the Files
Crucially, the vital information was not in the immediate customer interaction or superficial files. It was buried two document references deep within the company’s internal files. The models that read beyond the surface and uncovered this fact earned the full deal — worth over €4,583 MRR in real business terms. That’s a clear demonstration that reading and understanding deep documentation can be the decisive edge in negotiations.
Why This Matters for Business and Sports
In sports gear retail or any competitive industry, the ability of an AI to digest complex, layered information before making a decision could determine victory or defeat. Whether it’s assessing product quality, customer preferences, or internal operations, AI that thoroughly reads and understands all relevant data can identify hidden opportunities.
The Human Comparison
While chat demos often highlight AI’s conversational skills, this experiment shows a different metric: commitment to thoroughness and honesty. The models’ refusal of manipulative social engineering attempts indicates a level of discipline and integrity that is vital in high-stakes decisions — from closing deals to managing athlete health data or optimizing inventory.
What the Leaders Are Doing
The top performers in the experiment include:
- gpt-5.6-sol: Achieved a perfect score of 95, finding the buried fact and closing the deal.
- Kimi K3: The newcomer, scored 93, also closing the deal with the cleanest discipline.
Meanwhile, a more thorough but less disciplined participant, Opus 4.8, scored only 73 and failed to close the deal, illustrating that depth alone isn’t enough without discipline to act on insights.
The Implication for Your Business
Whether you’re managing a retail chain or a sports team, the key takeaway is straightforward: AI that reads deeply into your internal files — not just surface-level chats — can make the difference between closing or losing valuable opportunities. This ability to uncover hidden facts before making decisions is a measurable, critical property in AI agents, especially when stakes are high, and trust is paramount.
Try It Yourself
Enterprise users can test their own business scenarios through live wargames that simulate crises and decision-making under pressure. These experiments are designed to be fully transparent and reversible, allowing managers to see how their AI workforce performs in real-world conditions before deploying in actual operations.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
