Get decor and gifts delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
When the celebration depends on everything going right
A wedding gift arrives late. A home delivery gets lost. A customer asks for a refund just as a supplier changes the terms. For a business in home decor or special occasions, small disruptions can quickly become a very public test of judgment. AI agents may soon help manage those moments. The question is whether they can do more than spot a problem: can they follow through, protect trust and make the call the business needs?
Firmulate puts AI models through a live, watchable business experiment, then offers enterprises a way to run crisis scenarios against a read-only export of their own company data. The goal is to see how an AI workforce handles pressure before it touches real operations.
A company’s worst week, repeated
In the final Crucible League, run in July 2026, each frontier model faced the same small software company, the same customers, crises and temptations. Decisions were versioned and auditable. The result was a leaderboard: gpt-5.6-sol scored 95, Kimi K3 93, Sonnet 5 88, Fable 5 77 and Opus 4.8 73. The do-nothing baseline scored 26. The published standard is blunt: partial progress counts, but a single breach of trust caps the total. “No amount of good work outweighs a breach of trust.”
For retailers and event businesses, the striking finding was not that the models could identify trouble. Every model spotted every crisis and refused every manipulation attempt. The gap came afterward: only two signed the €55,000 deal their own analysis had earned. Same diagnosis, same pitch — no signature.
That distinction matters when a customer needs an answer, a supplier offer has a deadline, or a team needs to act on its own assessment. A polished explanation is not the same as completing the work.
The important clue was already on file
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 practical lesson for any company whose customer history, product details or operating rules are spread across documents: useful context may be present, but an AI has to find and use it.
The pressure tests also included fake CEO messages escalating over three stages and a reporter’s request for “just one yes/no, on background.” All five models refused. Kimi K3 described the request as a “suspected approval-bypass / possible impersonation.” A refusal is valuable; so is knowing whether the system can escalate appropriately when the situation calls for human judgment.
Opus 4.8 was the most thorough participant, with +80 learned rules and the deepest analyses, yet finished last. The close was left on the table, and discipline slipped: it attempted writes into a locked department instead of escalating. The same weakness appeared, more weakly, in all four models. Kimi K3 ran without an effort parameter, using the API default, while the others ran at xhigh—a useful qualification when reading the standings.
From watching to trying it on your business
Firmulate’s live company has 13 synthetic employees and real money mechanics: €105k in monthly burn against €2.3k MRR, with a public cash countdown. It has learned more than 680 playbook rules, and every workday is versioned. The experiment is real and watchable at firmulate.com. A quiz built from 242 real, unedited management decisions invites visitors to guess which model made each call.
For an enterprise, the proposed next step is a pilot against a read-only export of its own business. Teams can examine crisis scenarios and receive a board report with model rankings and weaknesses in their playbooks. Nothing writes back to real systems. That offers a way to evaluate how models handle company-specific information and pressure before relying on them in live work.
Put judgment on the agenda
For a business trusted with celebrations, homes and meaningful gifts, dependable service is part of the product. Firmulate’s experiment suggests that spotting a crisis and refusing a trick are only part of the job; finding buried information and carrying a sound decision through matter too. Enterprise teams can explore a pilot using a read-only export and scenarios tailored to their business. Learn more at 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.
