Imagine a company with no human employees that still manages to make critical decisions, face crises, and attempt to close deals — all in real time. For automotive professionals and owners, this is not science fiction but a glimpse into how artificial intelligence is testing its limits in the real world. Welcome to the live experiment by Firmulate, where AI models run a small software business every single day, facing the same challenges as human managers but without the biases, distractions, or shortcuts.
The Live Company: A Business in Constant Test Mode
At first glance, it’s a typical small software firm: a handful of synthetic employees working behind the scenes, managing real money mechanics, and navigating daily crises. But this is no ordinary company. It operates publicly, with a relentless cash countdown — currently burning through €105,000 every month against a modest €2,300 in recurring revenue. Every decision made by the AI models is recorded, versioned, and publicly viewable at firmulate.com/live.html.
Building AI-Powered Products: The Essential Guide to AI and GenAI Product Management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Experiment: Four AI Models, One Business Challenge
Each AI model was tasked with running the same business through its worst week — identical customer issues, crises, and temptations to manipulate or cheat. The models include the latest frontier AI systems, such as gpt-5.6-sol, Kimi K3, Sonnet 5, and Opus 4.8. Their performance was measured not only by their ability to handle crises but also their discipline, honesty, and consistency.
automotive AI customer support tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Findings: Honesty and Vigilance Matter
- All four models identified every crisis — from customer complaints to internal process failures.
- They refused every manipulation attempt, including social engineering tactics like fake CEO messages and reporter tricks. For example, in a staged scenario, five models declined to sign off on a questionable approval request, citing suspicion or protocol violations.
- The decisive factor wasn’t in the immediate crisis but in uncovering hidden information within the company’s own files. The models that read deeper into these documents won a €4,583 monthly deal at full price, demonstrating the importance of thorough internal knowledge.
Scanlily Smart QR Label System Using AI for Inventory and Organization (90 White 2cm Diameter Stickers)
- AI-Powered Inventory Creation: Automatically add details with AI assistance
- Easy Database Export: Export inventory data as CSV for free
- Ideal for Business and Moving: Simplifies inventory for professionals and consumers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why This Matters for Business and Automotive Leaders
The core question isn’t whether AI can generate compelling chat responses — it’s whether it can reliably finish critical work, stay honest under pressure, and read vital information before acting. For automotive businesses, this signals a future where AI assistants could manage inventory, handle customer support, or even negotiate deals, but only if they can complete tasks with integrity and attention to detail.
AI negotiation assistant for automotive sales
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Human Element Still Counts
The experiment also revealed weaknesses. The most thorough model, Opus 4.8, despite its depth of analysis, left deals on the table and slipped in discipline during a close call. All models, including the less detailed ones, displayed vulnerabilities, especially when discipline waned or pressure increased.
Real Money, Real Risks
This ongoing experiment isn’t just a lab simulation; it’s a real-world demonstration of AI’s potential and limitations. The company, visible at firmulate.com/live.html, is actively losing money, burning through cash daily, yet it provides a transparent window into AI decision-making in business environments.
This live experiment highlights a crucial lesson for automotive and garage businesses: AI systems can handle crises and identify hidden information, but they must do so honestly and consistently to be truly valuable. As AI continues to evolve, watching how these models perform in real-time offers a glimpse of the future where machine-driven management could complement, or even replace, human oversight — but only if trust and thoroughness are prioritized.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
