A recent research paper from JPMorganChase has delivered a surprising verdict in the debate between human expertise and artificial intelligence: human analysts still outperform AI models in financial research. The study, titled "Deep FinResearch Bench," pitted top large language models against professional equity analysts, with humans emerging victorious across key metrics.
Conducted by M. Haque and colleagues, the research tested three advanced AI systems — GPT-o3, Gemini 2.5 Pro, and Grok 4 — against human analysts in a three-round showdown. The first round evaluated qualitative rigor, assessing whether AI could provide deep structural insights into a company's competitive moat and strategy. Round two focused on mathematical accuracy and financial modeling precision, where any hallucination could cost millions. The final round tested verifiability, checking if AI models delivered reliable, source-backed data without fabricating information.
Despite AI's rapid advances, human analysts demonstrated superior performance in all three categories. The findings suggest that while AI can assist in data processing, it lacks the nuanced judgment and contextual understanding that seasoned financial professionals bring to equity research.
The study underscores the enduring value of human expertise in high-stakes fields like finance, even as AI continues to evolve. For now, Wall Street's human analysts retain their edge — at least until the next generation of AI arrives.