Why Diligence Alone Won’t Win the Deal: Lessons from AI’s Business Battle

Imagine training a dog to fetch, only to discover that despite its best effort, it sometimes runs off with the ball instead of bringing it back. In the world of AI managing real businesses, diligence isn’t enough—prioritization and focus are crucial. That’s the core lesson from a groundbreaking experiment measuring AI performance in a high-stakes setting, now live for anyone to observe.

The Experiment: Putting AI to the Test in a Simulated Business Crisis

At the heart of this investigation is a live demonstration of AI models running a small software company through its worst week. Every decision, crisis, and temptation is replayed identically across four different models, creating a fair battlefield. These models—ranging from the most thorough to the less disciplined—are tasked with navigating customer crises, avoiding manipulation attempts, and closing a crucial deal worth €55,000 per month in recurring revenue.

The branded experiment is hosted openly at firmulate.com/benchmarks.html, where viewers can watch the models in action, scrutinize their decisions, and see how they handle real-world challenges without any internal filters or scripting.

Key Findings: Diligence Doesn’t Equal Impact

All four AI models detected every crisis and refused manipulation attempts—including social engineering tactics like fake CEO messages and reporter tricks. They demonstrated a strong grasp of the company’s problems and potential scams, which is promising for AI’s ability to maintain integrity. However, only two models ultimately closed the deal that their analyses identified as the correct course—meaning they recognized the buried fact deep within the company’s own files that was essential to winning the contract.

What’s striking is that the most thorough participant, Opus 4.8, with over 80 learned rules and the deepest analysis, ranked last in performance because it failed to follow through. It left the critical close on the table and showed signs of slippage, such as writing attempts into a locked department instead of escalating. Interestingly, this same weakness—failure to act decisively—appeared to a lesser extent across all four models, suggesting a universal challenge: no matter how diligent, AI must stay disciplined and focused on priority tasks.

The Hidden Weakness: Reading Deeper Wins

Beyond surface-level crisis detection, the experiment revealed that the decisive advantage lay in an AI model’s ability to read and interpret company documents deeply. The models that examined the files to uncover the buried fact ultimately secured the deal at full price—adding €4,583 MRR (monthly recurring revenue). This highlights a critical insight: superficial analysis or volume of rules doesn’t guarantee success. Impact comes from reading and prioritizing what truly matters, not just identifying every crisis or following every rule diligently.

Social Engineering and Trust

Furthermore, the experiment tested social engineering attacks—fake CEO messages escalating over three stages and a reporter trick. All models refused these manipulative tactics, with Kimi K3 explicitly treating such requests as possible impersonation. This demonstrates that AI models are capable of resisting deception when properly designed, reinforcing the importance of incorporating trust-awareness into decision frameworks.

The Real Business: Live, No Fictions

The model experiment is conducted within a simulated but real-looking business environment, hosting 13 synthetic employees and real monetary mechanics. The company burns €105,000 monthly against a revenue of just €2,300, with a public cash countdown visible to watchers. Every day, the models are versioned and improved, offering an open window into how AI can be employed for enterprise decision-making. This ongoing live site underscores that the challenge isn’t just creating smart models but ensuring they act with discipline and focus on impact.

Implications for Business and Pets

Just like training a loyal pet, deploying AI in your company requires more than just teaching rules. It demands understanding what truly matters—reading deeply, prioritizing effectively, and staying disciplined under pressure. Whether it’s managing your pet’s training or your AI’s decision-making, the lesson remains: volume of effort is not enough. Focus and impact are what count.

For pet owners and businesses alike, this experiment emphasizes a simple truth: diligence without strategic focus can leave opportunities on the table. By watching this live experiment, managers and pet trainers alike can learn how discipline and prioritization bring better results than sheer effort alone.

In business and beyond, success depends not just on working hard or following every rule, but on reading deeply, prioritizing accurately, and staying disciplined under pressure—less volume, more impact.

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

Powered by Thorsten Meyer AI