Investment Bankers Give AI a Reality Check

A new benchmark has put the capabilities of top AI models to the test, specifically evaluating their ability to handle tasks that junior investment bankers perform on a regular basis. The assessment involved having GPT-5.4 and Claude Opus 4.6 generate content on various financial topics, including company valuations, mergers and acquisitions, and financial modeling. The AI outputs were then reviewed by 500 investment bankers, who were asked to rate their accuracy and usability for client-facing purposes.

The results were stark: not a single AI output was deemed suitable for client delivery. The bankers cited a range of issues, from imprecision in financial calculations to flat-out errors in data interpretation. While some of the AI-generated content showed flashes of brilliance, it was not enough to overcome the fundamental flaws that prevented it from being used in a real-world setting.

Despite the disappointing results, the study's findings are not entirely surprising. Many experts have long warned that AI models, despite their impressive capabilities, are not yet ready for prime time. The results of this benchmark serve as a reality check for the industry, highlighting the need for continued investment in AI research and development.

The implications of this study are significant. If top AI models cannot even meet the basic standards of accuracy and usability, it raises questions about the future of AI adoption in the financial industry. Will we see a shift towards more human-AI collaboration, where AI tools are used to augment human capabilities rather than replace them? Only time will tell.