Augmenting Fundamental Analysis with Large Language Models: A RAG-Based System for Generating Investor Briefs

2025-08-30 · Source: cs.CL updates on arXiv.org · Field: Finance & Economics — Capital Markets & Investment Management, FinTech & Digital Financial Services, Economic Analysis & Policy · Depth: Expert, extended

Summary

A Retrieval-Augmented Generation (RAG)-based Large Language Model (LLM) system was developed to augment fundamental analysis of companies, generating investor briefs by integrating diverse data sources. Utilizing the gpt-4o model, the system processes company reports, macroeconomic data (like US GDP and inflation), and U.S. SEC EDGAR filings, enriched with Kitchin business cycle knowledge. Over a four-week pilot, it analyzed nine diverse companies (Lululemon, NVIDIA, Energy Fuels, Coinbase, Corebridge Financial, Pinterest, Novavax, Amazon, Tesla) and produced automatic briefs. Evaluated by nine individual investors, the system demonstrated capabilities in cross-domain synthesis, advanced information processing (news ranking, thematic clustering, prompt-guided quantitative analysis), and qualitative reasoning for unforeseen events like martial law in South Korea. User feedback confirmed its primary value in saving time and aggregating vast amounts of information.

Key takeaway

For individual investors managing diverse portfolios, this RAG-based LLM system offers a powerful co-pilot, automating extensive data aggregation and synthesis. You can receive timely, multi-layered analytical briefs, freeing up your time for strategic decision-making. Consider exploring AI-augmented tools to enhance your analytical depth and efficiency, especially for processing unstructured financial data and macroeconomic indicators.

Key insights

A RAG-based LLM system can automate and deepen fundamental investment analysis by synthesizing diverse data and generating tailored investor briefs.

Principles

• Ground LLM outputs in source documents to ensure factual fidelity.
• Integrate diverse data (macro, company, news) for multi-layered analysis.
• Apply predefined heuristics for transparent, repeatable qualitative judgments.

Method

The system preprocesses diverse financial and macroeconomic data, then sends it via API to gpt-4o in a RAG-like regime. It generates daily analytical briefs, dynamically adjusting focus and prioritizing companies with significant activity.

In practice

• Implement RAG to reduce LLM hallucinations in financial analysis.
• Use vector databases and K-Means for news clustering and thematic analysis.
• Tailor LLM prompts to perform specific quantitative calculations.

Topics

• Large Language Models
• Retrieval-Augmented Generation
• Fundamental Analysis
• Investor Briefs
• Financial NLP
• Macroeconomic Data

Code references

• amueller/word_cloud

Best for: NLP Engineer, Research Scientist, Investor, Data Scientist, AI Scientist

Related on AIssential

• Augmenting Fundamental Analysis with Large Language Models: A RAG-Based System for Generating Investor Briefs — LLMs with RAG can automate investor brief generation from diverse financial and macroeconomic data.
• InvestPhilBench: A Multi-Layer Benchmark for Evaluating Large Language Model Procedural Reasoning in Expert Investment Philosophy — Large language models exhibit a significant procedural reasoning deficit in expert investment philosophy, often masked by fluent output.
• Büyük Dil Modellerinin Hafızasını Güçlendiren Yaklaşım:Retrieval-Augmented Generation (RAG) — RAG enhances LLMs by providing external, real-time context, reducing hallucinations and improving factual accuracy.
• Fin-RATE: A Real-world Financial Analytics and Tracking Evaluation Benchmark for LLMs on SEC Filings — LLMs struggle with multi-document, multi-entity, and temporal financial reasoning, revealing critical gaps in current evaluation.
• Time Series Augmented Generation for Financial Applications — Tool-augmented LLMs can achieve high accuracy and low hallucination in complex financial time-series analysis.
• Vectorless RAG - Local Financial RAG Without Vector Database | Tree-Based Indexing with Ollama — Vectorless RAG uses document structure and LLMs for retrieval, eliminating vector databases for structured data.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.