500 Bankers Assess AI Outputs with Mixed Results
Key Takeaways
- 1500 investment bankers review AI models for client tasks.
- 2None of the outputs were deemed client-ready.
- 3Over half of bankers would still use AI as a starting point.
A recent evaluation involved 500 investment bankers testing top AI models like GPT-5.4 and Claude Opus 4.6 on everyday tasks typically performed by junior staff. Despite their advanced capabilities, not a single AI output met the accuracy required for client delivery, showcasing significant limitations in current AI performance for critical applications in finance. The findings suggest while current AI models may not be ready for high-stakes use, their outputs could serve as preliminary tools for human bankers. This indicates a potential shift in how AI is integrated into financial workflows, driving evolution in AI’s role rather than outright replacement.