Do large language models displace or complement investment advisors? Evidence from Chinese wealth management firms

The emergence of Large Language Models (LLMs) has reignited debates about whether technology substitutes or complements human labor in knowledge-intensive professions. This study investigates LLM adoption among investment advisors in China's wealth management industry. Using a difference-in-differences framework with firm-level LLM adoption identified through recruitment text analysis, we examine effects on employment, skills, service quality, and pricing. Contrary to displacement predictions, LLM adoption significantly increases advisor headcount, driven by task reallocation: automation of routine analytical work frees advisors for high-value client interactions, expanding service capacity and stimulating hiring, particularly for junior and support roles. Firms raise soft skill requirements while demand for technical hard skills remains unchanged. Service quality improves substantially, with higher portfolio returns and client retention, yet advisory fees remain stable, suggesting that competitive pressures channel technological surplus to consumers. These findings challenge white-collar unemployment anxiety, showing that in relationship-intensive services with latent demand, AI-enabled scaling can dominate displacement.