FALLMAS: An LLM-Based Multi-Agent System for Automated Financial Analysis and Strategic Reporting

Abstract

Large language model-based Multi-Agent systems (LLM-MAS) latest success has proven remarkable versatility across several applications including domain-specific knowledge such as customs compliance and law. However, applying LLM-MAS to high-stakes domains like finance in particular, to financial analysis, which is a highly competitive task that requires a combination of effective strategy, sufficient knowledge and precise decision-making, is still rare and not fully proven. It is appealing to apply the emerging power of these Artificial Intelligence (AI) Agents in a competitive arena of financial advising in order to understand if they can outperform professional financial advisors. In this study, we introduce our Financial Advisor based on Large Language Models and multi-agent Systems (FALLMAS) as a fully operational LLM-MA system that integrates data extraction (e.g., from tax filings and web sources), financial ratio computation (e.g., Earnings Before Interest, Taxes, Depreciation, and Amortization (EBITDA), Return on Equity (ROE), gearing ratio), strategic SWOT analysis, and narrative report generation. Each autonomous agent specializes in a distinct task, such as financial metric mining, web exploration, news retrieval and strategic reasoning enabling effective analysis. This paper intends to provide interpretations into the current state of the art of LLM-based Multi-Agent Systems used in financial analysis. Furthermore, our LLM-based MA system for financial analysis is open-source, we invite researchers to make use of our results. The implementation will be released publicly upon acceptance.