Skip to content
AI IntelligenceApr 26, 2026AI Intelligence
Article

In a benchmark, 500 investment bankers reviewed outputs from AI models like GPT-5.4 and Claude Opus 4.6.

The sobering result: Not a single AI output was ready for client delivery – underscoring the continued weaknesses of AI in highly regulated, error-sensitive fields.

Data Cube AI EditorialSource: The Decoder
01

Source Brief

In a benchmark, 500 investment bankers reviewed outputs from AI models like GPT-5.4 and Claude Opus 4.6. The sobering result: Not a single AI output was ready for client delivery – underscoring the continued weaknesses of AI in highly regulated, error-sensitive fields.

02