AI agents are increasingly being positioned as tools for investment research, but their ability to retrieve even relatively simple financial data points can vary considerably. Is this primarily a model problem, or a data problem? In this literature review, we examine a new benchmark that assesses how accurately AI agents retrieve financial data when given web search and structured financial data tools. We then discuss the implications for investors building agentic research workflows and highlight datasets that could support similar applications.

LITERATURE

We review the January 2026 paper ‘FinRetrieval: A Benchmark for Financial Data Retrieval by AI Agents’ by Eric Y. Kim and Jie Huang of Daloopa. The paper introduces FinRetrieval, Daloopa’s benchmark for testing AI agents’ ability to retrieve specific financial values (e.g. a company’s quarterly revenue) across different model providers, tool configurations and reasoning modes.