Imagine running a furniture store or interior design firm where your entire management team is made up of AI models—yet the business still faces daily losses and tough choices. This isn’t science fiction; it’s the real-time experiment happening on Firmulate, where artificial intelligence is tested in the most extreme and transparent way possible.
The Experiment: AI Companies Living in Public
At the heart of this groundbreaking project is a small, simulated company managed entirely by AI models. Four of the world’s leading frontier AI models—ranging from GPT-5.6 to Sonnet 5—were tasked with running the same week of a business that faces typical crises: customer complaints, supplier issues, and the temptation to cut corners.
What makes this experiment extraordinary is its transparency. Every decision, every crisis response, and even the internal analysis of the AI models is publicly available, versioned, and auditable. The goal isn’t just to see if AI can mimic human decision-making, but whether it can do so honestly and effectively under pressure.
Key Findings: Honesty, Competence, and Gaps
- All four models identified every crisis and refused manipulation attempts, including social engineering tricks like fake CEO messages and reporter tricks. For example, Kimi K3 explicitly refused to approve suspicious requests, treating them as potential impersonation.
- Despite this, only two of the models managed to secure the €55,000 deal, which would have added €4,583 monthly recurring revenue (MRR). The others identified the opportunity but failed to follow through, leaving potential revenue on the table.
- The decisive weakness wasn’t in recognizing crises but in reading critical internal files. Models that examined deeper documentation uncovered a key fact buried two documents deep, which was instrumental in closing the deal.
- The experiment also measured discipline and consistency. Opus 4.8, the most thorough participant with over 80 learned rules, ultimately failed to close the deal because it slipped in discipline—writing attempts were contained in a locked department instead of escalated.
This experiment demonstrates that AI models can be honest and crisis-aware, but maintaining consistent, strategic discipline remains a challenge. It’s a stark reminder that AI’s decision-making quality hinges on more than surface-level interactions.
As an affiliate, we earn on qualifying purchases.
The Live Company: Real Money, Real Losses
The experiment isn’t just virtual. The live company managed by AI operates every business day, with 13 synthetic employees and real money mechanics. It burns €105,000 each month against a mere €2,300 monthly recurring revenue, with a public cash countdown displayed at firmulate.com/live.html.
The firm’s rules and decision processes are openly versioned daily, creating a transparent laboratory for observing AI’s capabilities and shortcomings in a real-world-like environment. This ongoing story offers a rare glimpse into how AI manages (or struggles to manage) actual business operations, including crises and ethical dilemmas.
What Interior Design and Furniture Businesses Can Learn
- The core concern isn’t whether AI can generate compelling chat responses, but whether it can see through complex crises and follow through on commitments.
- As businesses increasingly integrate AI into sales, support, or inventory management, understanding these decision capabilities becomes crucial.
- Trusting AI to handle sensitive decisions requires assurance that it recognizes and prioritizes internal insights—like the buried document fact—that can make or break deals.
The Takeaway: Building Trust in AI-Driven Business
For interior designers and furniture retailers, the message is clear: the future of automation isn’t just about quick replies or optimized logistics. It’s about building systems that can withstand the pressure of real crises, read vital internal data, and stay honest under temptation.
The Firmulate experiment offers a live window into this future—where AI companies are transparent about their daily struggles and reveal what it truly takes for AI to serve as a trustworthy decision-maker in business.
Watch it live: firmulate.com/live · Full results: firmulate.com/benchmarks.html
