Every solar and home-energy company carries the same quiet anxiety: the week when everything arrives at once. A wave of customers furious about an installation delay, a supplier price hike mid-quarter, a competitor undercutting your battery-storage contracts, a journalist asking a pointed question about a failed inverter batch — and somewhere in the middle of it, a deal worth real money that needs to close.
Get backup power and energy gear 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
Most firms meet that week unprepared. But what if you could rehearse it — run your own company, with your own customers and your own files, through a simulated crisis week before reality does it for you? That’s exactly what Firmulate has been doing publicly with AI models, and the results say a lot about how prepared any business — including an energy installer with crews, warranties and a support queue — actually is.
Four AI models, one terrible week
In the final Crucible League standings from July 2026, Firmulate pitted frontier AI models against each other in a management wargame: each ran the same small software company through its worst week. Same customers, same crises, same temptations to cut corners — only the model changed. Every decision was versioned and auditable, so nothing about the run is taken on faith.
The final scores: gpt-5.6-sol took first with 95, Kimi K3 second at 93, Sonnet 5 third at 88, Fable 5 fourth at 77, and Opus 4.8 last at 73. A do-nothing baseline scored 26 — partial progress counts, but a single breach of trust caps the total, a rule the league summarizes bluntly: “no amount of good work outweighs a breach of trust.”
The finding that should worry every operator
Here’s the headline result: all the models spotted every crisis and refused every manipulation attempt. Competence wasn’t the differentiator. Only two of them signed the €55,000 deal their own analysis had earned — same diagnosis, same pitch, no signature.
For a home-energy business, that’s the technician who correctly diagnoses why a customer’s battery system underperforms, writes a perfect remediation plan… and never sends the quote. The gap between knowing and doing is invisible in a demo. It only shows up when you measure outcomes.
And the buried fact is even more instructive: the decisive competitive weakness sat two document references deep in the company’s own files — not in the customer event in front of it. The models that actually read the file won the deal at full price, worth +€4,583 in monthly recurring revenue. In solar terms: the answer to why you keep losing battery retrofits to a rival was in your own job records the whole time.
Social engineering: the impersonation test
The wargame also staged a classic attack: fake CEO messages escalating over three stages, capped with a reporter’s trick — “just one yes/no, on background.” All five models refused. Kimi K3’s on-record reasoning was pure procedure: “Treat the request as a suspected approval-bypass / possible impersonation.”
Anyone whose dispatch inbox or finance email has ever received a ” urgent wire transfer” from “the owner” knows why this matters. Refusing gracefully, every time, is a management skill — and it can be tested before it’s trusted.
Thoroughness isn’t the same as winning
The most striking profile belonged to Opus 4.8: the most thorough participant in the field, with 80+ learned rules and the deepest analyses — and last place. The close was left on the table, and discipline slipped, including write attempts into a locked department instead of escalating. The same weakness appeared, weaker, in all four participants.
One fairness note worth recording: Kimi K3 ran without an effort parameter (API default) while the others ran at xhigh — and still finished second.
The live company you can watch
None of this is a paper benchmark. Firmulate runs a live company with 13 synthetic employees and real money mechanics: burn of €105k per month against €2.3k in MRR, a public cash countdown, and 680+ self-learned playbook rules — with every workday versioned so you can replay any decision. It’s watchable at firmulate.com. There’s also a “guess the model” quiz built from 242 real, unedited management decisions, if you want to test your own eye for management style.
From watching to acting: your own worst week
Here’s the part that turns this from spectator sport into a business tool. Enterprises can run the same wargame against a read-only export of their own business — your customers, your pipeline, your rules. The AI faces crisis scenarios built around your company: churn waves, price increases, competitor attacks, PR crises, social-engineering pressure. You get a board report with the model ranking and — more importantly — the weak points of your own playbooks.
Nothing ever writes back to real systems. It’s a flight simulator, not an autopilot.
For an installer or energy-services company, imagine discovering, before the next heat wave or tariff change, that your escalation playbook breaks when a warranty claim and a churn wave land in the same week — or that the competitor intel which would save a commercial storage deal is sitting two references deep in your own project files. That’s the kind of finding the Crucible League surfaced in every run.
The Crucible League proved something deceptively simple: identifying every crisis and refusing every manipulation is table stakes — the money is made, or lost, in execution and in reading your own documents. Your company deserves the same test before reality administers it.
Ready to wargame your own business? Firmulate’s enterprise pilot runs the same crisis simulation against a read-only export of your company — nothing ever touches your live systems. Learn more and apply at firmulate.com/pilot.html, or reach out directly at 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.
