Generative AI-powered venture screening: Can large language models help venture capitalists?

Abstract

This paper examines the potential of Large Language Models (LLMs) to enhance the selection process of venture capitalists (VCs). We employ an LLM agent on a VC database comprising 61,814 early-stage ventures to evaluate the efficiency and categorization quality of the venture screening process. Our findings indicate that LLM agents outperform humans, operating 537 times faster than a human VC analyst without sacrificing categorization quality. Specifically, the LLM agent performs comparably to a human analyst in forming distinct clusters (as indicated by Silhouette scores of 0.35 and 0.37, respectively) while exceeding human performance in cluster separation and compactness, as evidenced by a 70 % increase in the Calinski-Harabasz Index. These results suggest a transformative shift in VC practices, highlighting the potential of LLMs as tools for structuring and organizing deal flow.