LG AI Research developed a financial AI service that analyzes approximately 8,000 listed companies in the US and Korea, according to Lee Hwa-young, Executive Director of Business Development at LG AI Research, in an interview on the 24th. The AI system uses multiple specialized agents—an AI Economist for macroeconomic analysis, an AI Journalist for news, an AI Analyst for corporate disclosures, and a Financial Forecasting Agent to synthesize their outputs—to generate one-month company outlook scores (1-100) with explanatory text. Lee stated the service aims to reduce information gaps by providing retail investors access to analysis previously available only to institutional investors, with Kiwoom Securities serving as the first domestic client to deliver the AI-generated forecasts via its mobile trading platform.
LG AI Research Deploys Multi-Agent Architecture for Company Analysis
LG AI Research structured its financial AI service using multiple specialized agents that collaborate to analyze companies. Lee explained, "Just as people organize analyst teams to analyze companies by dividing roles, we made AI collaborate by dividing roles as well." The system includes an AI Economist that understands macroeconomic trends, an AI Journalist that analyzes news, an AI Analyst that examines corporate disclosures and financials, and a Financial Forecasting Agent that synthesizes these inputs into final forecasts. A separate verification AI checks for errors in company names and ticker codes, consistency between forecast scores and explanatory text, and alignment between industry trends and individual company outlooks. Lee stated, "It's a structure where AI predicts and another AI verifies, then feeds back for revision—AI performs large-scale verification that would be difficult for people to do repeatedly."
The service covers approximately 5,500 US companies and 2,500 Korean companies, excluding some firms and ETFs with insufficient data. Lee noted, "One model alone cannot properly analyze financial markets, so we adopted an agent structure where multiple AIs collaborate by dividing roles." Each company forecast is based on analysis of an average of approximately 12,000 news articles, which represents not just current-day articles but years of accumulated news, disclosures, and financial data.
LSEG and KOSCOM Supply Data While Kiwoom Securities Delivers Retail Service
London Stock Exchange Group (LSEG) and KOSCOM provide source data and serve as sales channels, while Kiwoom Securities is the first domestic client implementing the service for retail investors. KOSCOM and LSEG provide news, disclosures, stock price data and other information, which LG AI Research analyzes with AI to create forecast data. Kiwoom Securities will reconfigure this data to fit its mobile trading system (MTS) and provide it to investors.
Lee stated, "Currently in the market there are services that forecast the future, but there aren't many products that provide explanations of why they made those judgments along with the forecast." He added, "Contracts are being made as clients confirm a plus alpha compared to existing methods through backtesting." LG AI Research emphasized that explainable AI is crucial because even if AI predicts well, it is difficult to use in actual decision-making without presenting the basis for judgment. Lee noted, "Even within LG, when conducting demand forecasting or raw material price prediction projects, we received a lot of feedback saying 'We understand that AI predicts better, but it's hard to trust why it predicted that way.'" He added, "We confirmed that explanatory power is as important as accuracy in fields that predict the future, and we developed the explanatory text generation function together with the same awareness of this issue in finance."
AI Analyzes 12,000 News Articles Per Company but Excludes Short-Term Supply-Demand Factors
LG AI Research identified the AI's greatest strength as its overwhelming information processing scale. To create one company's forecast, the system analyzes an average of approximately 12,000 news articles, which includes not just same-day articles but years of accumulated news, disclosures, and financial data. The ability to simultaneously analyze large-cap, mid-cap, small-cap, and nano-cap companies was presented as a differentiator from existing research.
Lee acknowledged limitations, stating, "The current model is structured to forecast the future centered on company and industry fundamentals—it's difficult to predict supply-demand because the decision-making processes of institutions and foreigners are not publicly disclosed." He added, "The Korean market has greater influence from investor sentiment and supply-demand compared to the US, making prediction difficulty relatively higher, but AI has the advantage of quickly learning and adapting to new market environments."
LG AI Research Plans Supply Chain Analysis via Knowledge Graphs
LG AI Research plans to expand the service beyond individual company forecasts to supply chain analysis. The company is developing technology to build knowledge graphs based on disclosures, news, and corporate websites that connect suppliers, customers, and transaction relationships, visualizing supply chain structures and dependencies between companies.
Lee stated, "The goal is to enable general investors to utilize information and analytical capabilities that only professional investment institutions could access." He added, "Rather than AI replacing people's jobs, AI will play a role in analyzing more information neutrally to help rational decision-making."
FAQ
What companies does LG AI Research's financial AI service analyze?
LG AI Research's AI service analyzes approximately 8,000 listed companies, including approximately 5,500 US companies and 2,500 Korean companies. Some companies and ETFs with insufficient data are excluded from the analysis.
How does LG AI Research's multi-agent AI system generate company forecasts?
The system uses specialized AI agents that divide roles: an AI Economist analyzes macroeconomic trends, an AI Journalist analyzes news, an AI Analyst examines corporate disclosures and financials, and a Financial Forecasting Agent synthesizes these inputs. A verification AI then checks for errors and consistency, feeding results back for revision. Each company forecast is based on analysis of an average of approximately 12,000 news articles and years of accumulated data.
