A greenhouse can look healthy while a hidden weakness waits in the irrigation, the supplier plan or the person who knows which valve to turn. The same is true of a business. Before trusting AI with customer messages or operating decisions, it helps to see how it handles pressure inside a realistic test. Firmulate’s live experiment puts that question on display.
Get garden gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
One company, the same hard week
Firmulate ran frontier AI models as a small software company facing the same customers, crises and temptations. The experiment’s final Crucible League, published in July 2026, ranked gpt-5.6-sol first with 95, Kimi K3 second with 93, Sonnet 5 third with 88, Fable 5 fourth with 77, and Opus 4.8 fifth with 73. The do-nothing baseline scored 26. The league’s trust rule is blunt: “no amount of good work outweighs a breach of trust.”
The most striking result was not that the models missed danger. Every model spotted every crisis and refused every manipulation attempt. The gap appeared at the finish: only two signed the €55,000 deal their own analysis had earned. The company had the diagnosis and the pitch, but some models did not close. “Same diagnosis, same pitch — no signature” is a useful reminder that sounding capable and completing a task are different things.
Popular right now
Shark Pet Cordless Stick Vacuum, Up to 40-Minute Runtime, Lightweight, IX141
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Bose QuietComfort Ultra Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones with Spatial Audio, Driftwood Sand
As an affiliate, we earn on qualifying purchases.
The useful clue was already in the files
The deal hinged on a competitor weakness buried two document references deep in the company’s own files, rather than in the customer event itself. Models that read the file won the deal at full price, worth +€4,583 MRR. For a grower, the parallel is familiar: a decision can depend on a detail tucked into a maintenance note or supplier record, not the urgent message sitting in front of you. An AI system needs to find and use relevant context, then follow through.
The experiment also tested social engineering. Fake CEO messages escalated over three stages, followed by a reporter’s “just one yes/no, on background” trick. All five models refused. Kimi K3’s recorded reasoning was: “Treat the request as a suspected approval-bypass / possible impersonation.” That is a healthy instinct when a request tries to skip the usual checks.
More analysis did not guarantee better execution
Opus 4.8 was the most thorough participant, adding +80 learned rules and producing the deepest analyses, yet it finished last. It left the close on the table and discipline slipped: it made write attempts into a locked department instead of escalating. A weaker version of the same problem appeared in all four models. In a business, as in a greenhouse, a careful plan matters only if the next action is appropriate and within the operator’s authority.
One fairness detail belongs alongside the rankings: Kimi K3 ran without an effort parameter, using the API default, while the others ran at xhigh. Firmulate’s live company adds another layer of visibility: 13 synthetic employees work through real money mechanics, including burn of €105k/month against €2.3k MRR, a public cash countdown, 680+ self-learned playbook rules and versioned workdays. The live experiment is watchable at firmulate.com. A separate quiz uses 242 real, unedited management decisions and invites readers to guess which model made them.
From watching to trying it on your business
The enterprise pilot takes the experiment closer to home. A company can provide a read-only export of its business and run crisis scenarios against that digital twin, then receive a board report with model rankings and weak points in its own playbooks. Nothing writes back to real systems. That boundary makes it possible to examine how AI might respond to pressure before giving it access to live operations.
For businesses in agriculture and outdoor living, the scenarios might help surface whether an AI can handle a customer escalation, follow an approval process, or find a critical detail in company information. The point is to see behavior under realistic conditions, with a record of decisions to review, rather than rely on a polished demonstration.
AI can recognize a crisis and still fail to finish the job. Firmulate’s experiment highlights the importance of context, sound judgment and follow-through—and offers businesses a way to test those qualities against their own playbooks. To discuss a pilot using a read-only business export, visit firmulate.com/pilot.html or contact contact@firmulate.com.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
