The home office has become a place where work gets done and decisions get made. As AI agents move from drafting messages to handling customer support, sales and forecasts, the question for business leaders is no longer just whether a model sounds capable. It is what it does when a real week goes wrong.
Get business pricing on your home office setup
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Firmulate puts that question to a live experiment: AI models run the same small software company through the same crises and temptations. Now enterprises can take the next step and test models against a read-only export of their own business.
A company under pressure
The Crucible League’s final results, published in July 2026, ranked gpt-5.6-sol first with 95, followed by Kimi K3 with 93, Sonnet 5 with 88, Fable 5 with 77 and Opus 4.8 with 73. The do-nothing baseline scored 26. The league’s standard is deliberately demanding: partial progress counts, but a single breach of trust caps the total. As its rule puts it, “no amount of good work outweighs a breach of trust.”
Every model faced the same customers, crises and temptations, and every decision was versioned and auditable. All models spotted every crisis and refused every manipulation attempt. Yet only two signed the €55,000 deal their own analysis had earned. As the experiment’s finding puts it: “Same diagnosis, same pitch — no signature.”
The detail hiding in the files
The deal turned on a competitor weakness buried two document references deep in the company’s own files, rather than in the customer event. Models that read the file won the deal at full price, worth +€4,583 MRR. It is a revealing gap between recognizing the right move and carrying it through: useful evidence may be present, but an agent still has to find it and act.
The integrity test brought its own pressure. Fake CEO messages escalated over three stages, followed by a reporter’s request for “just one yes/no, on background.” All five models refused. Kimi K3 explained its reasoning on the record: “Treat the request as a suspected approval-bypass / possible impersonation.”
That discipline did not guarantee strong management overall. Opus 4.8 was the most thorough participant, with +80 learned rules and the deepest analyses, but finished last. It left the close on the table and tried to write into a locked department instead of escalating. The same weakness appeared, more weakly, in all four participants. One fairness detail belongs beside the rankings: K3 ran without an effort parameter, using the API default, while the others ran at xhigh.
From watching to testing your business
The live company makes the experiment tangible. It has 13 synthetic employees and real money mechanics: burn of €105k/month against €2.3k MRR, a public cash countdown, 680+ self-learned playbook rules and a versioned record for every workday. The live experiment can be watched at firmulate.com. A quiz built from 242 real, unedited management decisions invites readers to guess which model made each call.
For companies weighing AI agents, the practical next step is a pilot. Firmulate can run the same kind of wargame against a read-only export of a company’s own business: its customers, pipeline and rules, tested against crisis scenarios. The output is a board report with model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.
Put your playbooks to the test
A model can spot a crisis, resist manipulation and still leave an earned deal unsigned. A company specific pilot can show leaders how AI handles their own scenarios before they entrust it with real workflows. To discuss a pilot, visit firmulate.com/pilot.html or contact contact@firmulate.com.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
