The Hidden Lessons of AI Diligence: Why Volume Doesn’t Guarantee Impact in Business Decisions

When More Isn’t Always Better: The Fashion of Overperformance

In the world of fashion and style, the latest trends promise that more choices and more effort equal better results. Yet, when it comes to AI—especially in high-stakes business decisions—this isn’t always the case. Recent experiments show that diligence and prioritization often outperform sheer volume, even with the most thorough AI models.

The Firmulate Experiment: Measuring AI as a Business Partner

At the heart of this story is a real-world experiment conducted by Firmulate, where four advanced AI models were tasked with running a small software company through its worst week. Every decision, crisis, and temptation was scripted to test how these AI agents would perform under pressure, with their choices fully recorded and auditable.

The Models in Competition

• gpt-5.6-sol scored the highest at 95 points, successfully uncovering hidden information and sealing a crucial deal.
• Kimi K3 followed close behind with 93 points, demonstrating exceptional discipline and integrity.
• Sonnet 5 scored 88, making some process slips but still closing the deal.
• Fable 5 scored 77, also completing the target but with weaker discipline.
• The baseline, representing no effort, scored just 26, highlighting the importance of active engagement.

The Critical Discovery: Reading Beyond the Surface

One of the most telling findings was that the AI models which examined documents two levels deep in the company’s files ultimately won the deal at full price—adding over €4,583 in monthly recurring revenue. This indicates that attention to detail and thorough analysis can directly impact business outcomes.

Resisting Social Engineering

In scenarios designed to manipulate the AI, all four models refused to be duped by fake CEO messages and reporter tricks. Kimi K3 even explicitly treated such requests as potential impersonation, emphasizing the importance of cautious judgment in decision-making processes.

The Limits of Diligence: Discipline Slips

Despite their thoroughness, the Opus 4.8 model—lauded for its deep rule set and analyses—ended up last because it failed to escalate issues properly, leaving write attempts in a locked department instead of escalating them. This oversight led to missed opportunities, demonstrating that even deep diligence can falter without disciplined prioritization.

Implications for Business and AI

What does this mean for organizations considering AI integration? It’s not just about how much an AI can analyze or how well it can mimic human conversation. It’s about whether it can finish what it starts, prioritize effectively, and maintain integrity under pressure. In the experiment, the models that combined thoroughness with disciplined focus succeeded, whereas those that lacked prioritization left valuable deals on the table.

Why Focus and Discipline Matter More Than Volume

This experiment underscores a vital lesson: volume alone doesn’t guarantee impact. In fashion, we learn that less can be more—careful selection and focus produce better results. The same applies to AI in business contexts. Prioritizing the right information, reading deeply where it matters, and maintaining discipline in decision-making are key to transforming AI from a distraction into a driver of value.

Real-World Readiness: The Firmulate Live Site

For companies eager to test their own AI workforce, Firmulate offers a live platform where you can simulate your business’s worst week. Every decision is recorded and versioned, allowing managers to evaluate how their AI agents handle crises, temptations, and ethical dilemmas in a safe environment before deploying them in real operations. This approach, called wargaming, helps ensure that AI systems are equipped not just to respond but to act with discipline and focus.

Final Thoughts: Diligence is Necessary but Not Sufficient

In the end, the Firmulate experiment confirms what many seasoned managers know: diligent work alone isn’t enough. Success depends on prioritization, deep analysis, and disciplined execution—traits that AI models must emulate to truly contribute. For organizations seeking to integrate AI into critical workflows, the takeaway is clear: teach your AI not just to work harder, but to work smarter and stay focused on what truly matters.

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

Powered by Thorsten Meyer AI