AI’s True Test: Can It Finish the Job When It Matters Most?

Firmulate — Four AI Models Ran the Same Company Through Its Worst Week. Only Two Finished the Job. — Live on firmulate.com.

In the world of interior design and furniture, we often focus on aesthetics and ideas. But when it comes to AI in business, the real question isn’t how well it chats — it’s whether it can actually finish the work, especially when stakes are high. Imagine if your AI assistant could spot every crisis and resist every temptation, yet still fail to close the deal. That’s exactly what a recent experiment reveals about AI’s true strengths and weaknesses.

Testing AI in the Crucible of Business

Recently, four leading AI models were put through a rigorous test: managing a small software company facing its worst week. The goal was straightforward — see if the AI could navigate crises, resist manipulation, and close a €55,000 deal based on its own analysis. Every decision was currency, every step observed and auditable. This wasn’t a chat simulation; it was a real-world business scenario with real money at stake.

The Models and Their Scores

gpt-5.6-sol 95: Achieved the highest score, identifying the buried fact in the company’s files and closing the deal.
Kimi K3 93: A newcomer with the most disciplined approach, also successfully closed the deal.
Sonnet 5 88: Managed to close the deal but with some process slips.
Fable 5 77: Demonstrated strong rule discipline but failed to execute the deal, leaving money on the table.

What’s noteworthy is that all models demonstrated awareness; they identified every crisis and refused manipulation attempts, including staged social engineering tactics. No chat demo can reveal whether an AI has the grit to see a project through to completion. The real difference emerged in their ability to act on the insights they gathered.

The Hidden Weakness

Interestingly, the decisive advantage for the top performers lay in reading deeper company files, two document references down. Those that read the full context and understood the company’s own data closed the deal at full price, adding over €4,500 in monthly recurring revenue. This shows that surface-level chat capabilities are only part of the story; true business acumen requires digging deeper into data and context.

Resisting Manipulation and Staying Honest

Beyond crisis management, the experiment tested whether models would be fooled by manipulative tactics, such as fake CEO messages or reporter tricks. All five models refused these social engineering attempts, with Kimi K3 explicitly reasoning that such requests might be impersonation or approval-bypass attempts. This highlights that AI’s resistance to manipulation isn’t just about avoiding errors; it’s about maintaining integrity under pressure.

AI Builders: Making The Decisions That Turn AI Code Into Real Software

As an affiliate, we earn on qualifying purchases.

What It Means for Your Business

For interior designers or furniture brands contemplating AI tools, the takeaway is clear: the ability to generate convincing chat responses is not enough. When integrating AI into critical workflows — whether managing client projects, supply chains, or negotiations — it’s essential to evaluate whether the model can actually complete tasks, read relevant data deeply, and resist shortcuts or manipulations.

The Performance Gap

The experiment’s results underscore an often-overlooked aspect: the gap between apparent competency in conversation and real-world execution. The most thorough participant, Opus 4.8, demonstrated the deepest analysis but faltered at the final step, leaving the deal unclosed. Conversely, models that prioritized disciplined reading and honest decision-making succeeded in closing the deal.

Implications for AI Adoption

These findings challenge the common assumption that chat quality equates to reliability. For businesses aiming to use AI for operational decisions, trustworthiness, and execution strength are the true measures of worth. Asking AI to perform complex, high-stakes tasks is a test — and only some models pass.

AI for Real Companies: A Practical Guide to Smarter Systems and Stronger Profits

As an affiliate, we earn on qualifying purchases.

How to Wargame Your AI Workforce

To prepare your business for AI integration, consider running simulated scenarios similar to this experiment. Firmulate offers an interactive platform where you can test your AI models against realistic crises, observe their decision-making, and identify weaknesses before deploying them in real work. This approach helps ensure that when your AI faces real pressures, it can deliver results — not just generate plausible responses.