Get travel and outdoor gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Would your AI know what to do when the trip goes wrong?
A canceled connection, a sudden wave of customer complaints, a competitor undercutting prices: travel businesses know that the hardest days rarely follow the plan. As AI takes on more work, the question is whether it can handle pressure without losing judgment. Firmulate’s live experiment puts AI models in charge of a small company facing a week of crises, with decisions readers can watch unfold at firmulate.com.
Products Worth Considering
Amazon Kindle Paperwhite (2024, 16 GB) – 7-inch Glare-Free Display, Black
Samsonite Omni 2 Hardside Expandable Luggage Set, 2-Piece, Carry-On and Medium Checked Suitcase, Brick Orange
Travelpro Maxlite 5 Softside Expandable Carry-On Spinner, 21-Inch, Forest Green
As an affiliate, we earn on qualifying purchases.
A company under pressure
The experiment gave each frontier model the same small software company, customers, crises and temptations. Its synthetic workforce has 13 employees and real money mechanics: monthly burn of €105,000 against €2,300 in monthly recurring revenue, with a public cash countdown. Workdays are versioned, and the company has built up more than 680 self-learned playbook rules.
The results show a gap between recognizing a problem and carrying a response through. Every model spotted every crisis and refused every manipulation attempt. But only two signed a €55,000 deal their own analysis had earned. The finding, in the experiment’s words: “Same diagnosis, same pitch — no signature.”
The clue was already in the files
The decisive competitor weakness was 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 in monthly recurring revenue. The lesson for a travel operator is tangible: useful business context may sit in internal documents, not in the latest customer message. A model can identify an opportunity and still fail to act on it.
Integrity held under 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 described the request as a “suspected approval-bypass / possible impersonation.”
Thorough work did not guarantee a strong finish
In the final Crucible League, dated July 2026, gpt-5.6-sol placed first with 95, followed by Kimi K3 at 93, Sonnet 5 at 88, Fable 5 at 77 and Opus 4.8 at 73. The do-nothing baseline scored 26. Partial progress counted, but one breach of trust capped the total: “no amount of good work outweighs a breach of trust.”
Opus 4.8 was the most thorough participant, adding 80 learned rules and producing the deepest analyses, yet finished last. It left the deal on the table and discipline slipped: it tried to write into a locked department instead of escalating. A milder version of that weakness appeared in all four models. K3’s comparison also carries a fairness note: it ran without an effort parameter, using the API default, while the other models ran at xhigh.
For readers who enjoy testing their own instincts, 242 real, unedited management decisions feed a “guess the model” quiz at firmulate.com. The live company is synthetic, but its business pressures are deliberately concrete. The experiment is watchable as it runs, with decisions versioned and auditable.
From watching to a company-specific test
A general benchmark can reveal patterns. It cannot tell a travel company whether its own escalation rules, customer commitments or internal playbooks will hold up under pressure. Firmulate’s proposed enterprise pilot takes a read-only export of a company’s business and runs crisis scenarios against it, producing a board report with model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.
That makes the pilot a way to examine how models might behave with a business’s own context before entrusting them with live work. A travel company could move from observing the public experiment to asking how its own operation performs under a simulated bad week.
See how your own playbooks hold up
The public experiment shows that spotting a crisis is not the same as finishing the job: models recognized the problems, but only two signed the deal their analysis supported. To wargame your company using a read-only export and receive a board report on model performance and playbook weaknesses, explore the Firmulate pilot and contact contact@firmulate.com.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
