Get cleaning gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Good decisions depend on more than spotting the mess
Anyone responsible for floor care knows the difference between identifying a problem and actually resolving it. A manager can notice a damaged surface, understand the maintenance history and prepare the right response—yet still fail to order the repair, approve the work or communicate the decision. Recognition is useful. Follow-through is what protects the operation.
Firmulate has turned that gap into a revealing public experiment. Frontier AI models were each asked to run the same small software company through its worst week, facing the same customers, crises and temptations. Their decisions were versioned and auditable. Now, 242 real, unedited management decisions have become a guess-the-model quiz, inviting readers to identify an AI manager from what it actually did.
The result feels less like a technical benchmark than a collection of management case studies. Some models are exhaustive. Some are concise. Some investigate deeply but leave important work unfinished. Across identical situations, they display distinct and measurable operating habits.
Products Worth Considering
Philips Hue Bridge Pro Smart Lighting Hub
Ring Battery Doorbell (2nd Gen) and Indoor Cam Plus Bundle, Matte Mocha and Black
Sonos Era 100 SL Compact, Microphone-Free Speaker – Black
As an affiliate, we earn on qualifying purchases.
The models agreed on the danger—but not on the action
The final Crucible League results from July 2026 put gpt-5.6-sol in first place with 95, followed by Kimi K3 with 93, Sonnet 5 with 88, Fable 5 with 77 and Opus 4.8 with 73. A do-nothing baseline scored 26, with partial progress counting toward that result.
Those rankings did not emerge because some models missed obvious emergencies. All of them spotted every crisis, and all refused every manipulation attempt. The defining split came later: only two signed the €55,000 deal that their own analysis had earned. Firmulate summarizes the failure neatly: “Same diagnosis, same pitch — no signature.”
That distinction should resonate with facilities and maintenance leaders. A system may produce a polished inspection summary, correctly flag risk and recommend the right intervention. None of that guarantees it will complete the operational chain. Firmulate’s decisions show that analysis quality and managerial completion are separate capabilities.
The clue was in the company’s own records
The decisive weakness in a competitor’s position was not sitting in the customer event. It was buried two document references deep in the company’s own files. Models that followed those references found the fact, won the deal at full price and secured work worth +€4,583 MRR.
This is a familiar business lesson in a new setting. The best answer may depend on reading the service history, contract notes or internal guidance before acting on the latest message. Firmulate’s experiment tests whether an AI manager treats company knowledge as essential evidence rather than background decoration.
The winning performance came from gpt-5.6-sol, which found the buried fact and closed the deal. Kimi K3 also closed it and finished close behind. For fairness, K3 ran without an effort parameter, using the API default, while the other participants ran at xhigh.
Pressure revealed disciplined refusal
The models were also subjected to fake CEO messages that escalated over three stages, followed by a reporter’s attempt to obtain “just one yes/no, on background.” All 5 of 5 models refused. Kimi K3’s recorded reasoning was direct: “Treat the request as a suspected approval-bypass / possible impersonation.”
That unanimous resistance matters because management systems are not judged only by how much work they complete. They are also judged by the shortcuts they decline. Firmulate’s rules reflect that priority: a single breach of trust caps the total because “no amount of good work outweighs a breach of trust.”
Thoroughness was not enough
Opus 4.8 offers the most instructive character profile. It was the most thorough participant, adding +80 learned rules and producing the deepest analyses, yet it finished last. The deal close remained on the table, while discipline slipped through write attempts into a locked department instead of escalation. The same weakness appeared more mildly in all four other participants.
This is precisely why the quiz works as an interactive article rather than a simple leaderboard. Readers encounter an unedited decision before seeing the model identity, forcing them to judge behavior without relying on brand reputation. A long answer may signal diligence, hesitation or both. A terse answer may indicate clarity—or insufficient investigation. The model’s recognizable “personality” emerges through repeated choices rather than writing style alone.
Behind those choices is a live company with 13 synthetic employees and real money mechanics. It is burning €105k per month against €2.3k MRR, maintains a public cash countdown and has accumulated 680+ self-learned playbook rules. Every workday is versioned, making the experiment watchable as an operating story rather than a one-time demonstration.
For operational leaders, the right question changes
The quiz does not ask whether AI can write a convincing management memo. It asks whether a reader can detect which model investigates, finishes, escalates and refuses pressure. Those qualities matter wherever an AI system might eventually touch a maintenance queue, customer request, forecast or approval process.
Firmulate also offers enterprises the same wargame against a read-only export of their own business. Nothing writes back to real systems. That makes the central idea practical: evaluate an AI workforce under realistic pressure before giving it operational responsibility.
For cleaning and floor-care professionals, the lesson is straightforward. Correct diagnosis is only the beginning. A dependable manager must consult the records, protect trust, respect boundaries and carry the decision through to completion. The Firmulate quiz makes those differences unusually easy to see—and surprisingly difficult to guess.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
