AI-Augmented Financial Advisors: How AI Impacts Financial Firms in Client Portfolio Performance and Advisor Productivity Versus Human Advisors

Presenter: Crystal Chen

Faculty Sponsor: Zaur Rzakhanov

School: UMass Boston

Research Area: Finance

Session: Poster Session 2, 11:30 AM - 12:15 PM, Auditorium, A46

ABSTRACT

Human biases, such as overconfidence, confirmation bias, recency bias, and loss aversion, can distort judgment and lead to systematic deviations from rational asset valuation. Artificial intelligence (AI) has been proposed as a tool to mitigate these biases by systematically aggregating and analyzing extensive financial and macroeconomic data without emotional distortions. This study investigates the level of agreement and performance between AI-generated investment recommendations and those made by human analysts across U.S. publicly traded firms. Utilizing a sample of twelve companies classified by size (large, mid, and small), the study gathers buy, hold, or sell recommendations from both AI systems and human analysts. The agreement between AI and human recommendations is measured using Cohen’s Kappa for pairwise comparison and Fleiss’ Kappa for overall group agreement. Portfolio performance is evaluated by constructing equal-weighted portfolios for each recommendation source and size category. Risk-adjusted returns are measured using the Sharpe ratio over 1, 2, and 3-month periods. Additionally, the study examines whether disagreement between AI and human recommendations is more pronounced among smaller firms, where there is more information asymmetry and limited analyst coverage. Through analysis of agreement and portfolio performance evaluation, this research aims to determine whether AI systems align with, diverge from, or outperform traditional human analysts in short-term investment contexts.

RELATED ABSTRACTS