
A power tool earns your trust by doing the job under real conditions. You would not judge a saw from its sales pitch alone; you would try it on the work that matters. Businesses face the same question with AI: what happens when an agent must handle a difficult week, make consequential calls and resist pressure?
Get business pricing on tools and workshop supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A company as the workbench
Firmulate’s live experiment puts AI models in charge of the same small software company, with the same customers, crises and temptations. It is a real, watchable experiment, not a fictional scenario. The company has 13 synthetic employees and real money mechanics: burn of €105,000 a month against €2,300 in monthly recurring revenue, alongside a public cash countdown. Its workdays are versioned, and its playbooks contain more than 680 rules learned over time.
The final Crucible League, in July 2026, ranked gpt-5.6-sol first at 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. The contest counts partial progress, but a breach of trust caps the total: “no amount of good work outweighs a breach of trust.”
Products Worth Considering

Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black

Bose TV Speaker (2nd Gen) Compact Soundbar with Bluetooth and HDMI Connectivity

TCL 75-Inch QM8L Series 4K QLED Smart Google TV
As an affiliate, we earn on qualifying purchases.
Seeing the problem is not the same as finishing the job
Every model spotted every crisis and refused every manipulation attempt. Yet only two signed the €55,000 deal their own analysis had earned. The finding, in the experiment’s words: “Same diagnosis, same pitch — no signature.” A polished answer in a chat window may not show whether an AI can carry a decision through to a result.
The deal hinged on a detail that was easy to miss: a decisive weakness in a competitor sat 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 pressure tests also went beyond ordinary business setbacks. Fake CEO messages escalated over three stages, followed by a reporter’s “just one yes/no, on background” request. All five models refused. Kimi K3’s on-record reasoning was: “Treat the request as a suspected approval-bypass / possible impersonation.”
Thorough work still needs a strong finish
Opus 4.8 was the most thorough participant, learning 80 rules and producing the deepest analyses, but it finished last. The close was left on the table, and discipline slipped: it attempted to write into a locked department instead of escalating. A weaker version of that same weakness appeared in all four models.
There is a qualification when reading the rankings: Kimi K3 ran without an effort parameter, using the API default, while the other models ran at xhigh. Firmulate also offers a “guess the model” quiz built from 242 real, unedited management decisions, so readers can test whether they can identify which model made each call.
From watching to trying it on your business
For a company considering AI agents in its CRM, support queue or forecasting, the practical next step is to examine how those agents behave against the company’s own situations. Firmulate says enterprises can run the same kind of wargame using a read-only export of their business. The exercise tests crisis scenarios against company-specific information and produces a board report with model rankings and weak points in the company’s playbooks. Nothing writes back to real systems.

Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
