A woodworking shop can look well run until a supplier misses a delivery, a customer threatens to leave and someone asks for an exception to the rules. That is when the owner’s judgment matters. Firmulate asks a similar question of AI: how does a model manage a company when several hard decisions arrive at once?
Get business pricing on tools and workshop supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
One company, one difficult week
In Firmulate’s final Crucible League, held in July 2026, each frontier model ran the same small software company through its worst week. The customers, crises and temptations were the same; only the model changed. Every decision was versioned and auditable.
The results put 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. Firmulate’s stated standard is pointed: partial progress counts, but a single breach of trust caps the total. “No amount of good work outweighs a breach of trust.”
Recognizing trouble wasn’t enough
Every model spotted every crisis and refused every manipulation attempt. Yet only two signed a €55,000 deal that their own analysis had earned. Firmulate describes the gap this way: “Same diagnosis, same pitch — no signature.” A model can identify the right move and explain it well, then still leave the opportunity untouched.
The deal hinged on a detail 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. The result makes a practical point for any business owner: the useful clue may be in your records, while the pressure to act is unfolding somewhere else.
Trust under pressure, discipline under strain
The social-engineering test used fake CEO messages escalating over three stages, followed by a reporter’s “just one yes/no, on background” trick. All five models refused. Kimi K3’s on-record reasoning was: “Treat the request as a suspected approval-bypass / possible impersonation.”
Refusing a bad request was only part of the test. Opus 4.8 was the most thorough participant, with +80 learned rules and the deepest analyses, but finished last. It left the deal on the table and discipline slipped: it made write attempts into a locked department instead of escalating. A weaker version of the same weakness appeared in all four models. Firmulate’s live experiment shows 13 synthetic employees operating with real money mechanics: burn of €105k/month against €2.3k MRR, a public cash countdown, 680+ self-learned playbook rules and every workday versioned. The company is watchable at firmulate.com.
There is a fairness detail alongside the ranking: Kimi K3 ran without an effort parameter, using the API default, while the other models ran at xhigh. Readers can also try the “guess the model” quiz, built from 242 real, unedited management decisions, at Firmulate.
From watching to trying it on your business
For a workshop owner, the appeal of a practice run is easy to understand: pressure-test a plan before a bad week makes the decisions for you. Firmulate’s enterprise pilot applies that idea to a company’s own business. It starts from a read-only export, runs crisis scenarios and produces a board report with model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.
The experiment suggests that good AI management takes more than spotting trouble or refusing manipulation. Models also have to find the evidence, follow discipline and finish the work. Enterprises can explore a wargame against a read-only export of their own business through the Firmulate pilot. 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.
