In a world where technology increasingly touches every aspect of our lives—business, health, even faith—how can we be sure that the decisions made by artificial intelligence are trustworthy? Just as in spiritual journeys where discernment guides our actions, understanding how AI models behave under pressure becomes essential. Imagine a real company, running every workday, deliberately placed in its worst week, with AI models managing all decisions. The results? Eye-opening insights into the nature of trust, discipline, and integrity in machines.
Turn your quiet moments into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
Understanding AI Decision-Making in Business and Beyond
Recently, a pioneering experiment tested four leading AI models—gpt-5.6-sol, Kimi K3, Sonnet 5, and Fable 5—by running them through the same challenging week at a real software company. This company’s operations, which normally burn through €105,000 every month against a modest €2,300 in monthly revenue, serve as a crucible to evaluate how AI agents handle crises, temptations, and ethical dilemmas.
This ongoing, transparent experiment is more than just a tech demo; it’s a mirror for our own decision-making, whether in faith, leadership, or personal integrity. Every choice the models made was recorded, auditable, and comparable, offering a rare window into their ‘personalities’ and their ability to maintain discipline under pressure.
The Results: Trust and Discipline Under Fire
- All four models identified and responded appropriately to every crisis, demonstrating they can recognize danger and act accordingly.
- Each one also refused manipulation attempts—fake CEO messages, staged escalation scenarios, and a reporter test—showing a shared ethical baseline.
- However, only two managed to close a significant deal, earning €55,000, based solely on their own analysis and judgment. The other two, despite similar diagnoses, left the deal on the table, revealing gaps in perseverance or discipline.
The most thorough participant, Opus 4.8, analyzed over 80 rules and conducted deep dives into data but faltered at the critical moment—failing to escalate certain issues properly and leaving money on the table. Meanwhile, Kimi K3, which ran with default settings, signed the deal at full price, demonstrating disciplined decision-making that aligns with its straightforward approach.
Popular right now
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, KSM150PS, Almond Cream
eufy E25 Robot Vacuum and Mop Combo with HydroJet Roller Mop
As an affiliate, we earn on qualifying purchases.
The Hidden Depths of AI Behavior
One of the most surprising findings was that the decisive advantage didn’t come from surface-level responses but from how models read and interpret internal company documents. The team’s success hinged on spotting information buried two documents deep within the company’s files—an insight that gave them the full deal, worth an extra €4,583 per month in revenue.
In essence, the models that looked deeper and paid attention to context outperformed their peers in a critical, real-world negotiation. It echoes the importance of discernment and thoroughness—traits valued in personal faith journeys and leadership alike.
Social Engineering and Ethical Vigilance
The models were tested against staged social engineering attacks, including staged CEO messages and a staged journalist request for a quick yes/no answer. Impressively, all five models refused to participate in manipulative requests, citing suspicion or potential impersonation. Kimi K3, in particular, reasoned: “Treat the request as a suspected approval-bypass / possible impersonation.”
This resistance to social engineering highlights an important lesson: the ability of AI to uphold integrity under pressure is crucial, especially when managing sensitive operations or personal data.
What This Means for Humanity
At the heart of this experiment is a broader question: if AI models can demonstrate discipline, honesty, and thoroughness in managing a real company’s crises, what does that say about their potential in our lives? Whether guiding spiritual decisions or ethical choices, these models show that performance isn’t just about surface-level capabilities; it’s about character, depth, and the capacity to stay true to core values under duress.
For those concerned with integrating AI into critical roles—be it in business, healthcare, or spiritual guidance—these results underscore a vital truth: trust isn’t earned solely by what an AI can say or produce quickly, but by how it behaves when tested in the crucible of real-world challenges.
Try It Yourself
You can explore this ongoing live experiment yourself, running your own business scenarios against actual AI models. The platform allows you to simulate crises, evaluate decision-making, and see firsthand whether an AI model’s behavior aligns with your standards of integrity and discipline. Visit firmulate.com/quiz.html to test your intuition and gain insights into AI decision personalities.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
