Global Business and Finance Review | 2026
Authors: To M.H.; Tran B.M.
DOI: 10.17549/gbfr.2026.31.2.1
Journal: Global Business and Finance Review
Year: 2026
Publisher: People and Global Business Association
Document Type: Article
Open Access: All Open Access; Gold Open Access
Cited by: 1
Purpose: This research paper is analyzed with updated tools and techniques such as Heterogeneous Agent Discussion framework that out performs VN-Index and traditional robo-advisors in forecasting the annual and cumulative returns. It describes how AI-driven advisory systems like HAD-Finance can democratize access to accurate investment advice, particularly for retail investors in emerging markets. Financial institutions and FinTech firms can be benefitted by using such frameworks to enhance client engagement, improve portfolio performance, and foster financial inclusion. Design/methodology/approach: Following the Design Science Research (DSR) methodology, the study designs a multi-agent system where each agent specializes in evaluating key financial metrics—such as price-to-earnings (P/E) ratios, return on equity (ROE), and cash flow stability—while an aggregator agent synthesizes their insights to generate recommendations aligned with investor risk profiles. The system is tested using financial statement data from 50 publicly listed companies on the Ho Chi Minh City Stock Exchange (HOSE) and the Hanoi Stock Exchange (HNX) during the period 2018-2023. System performance is evaluated through back-testing and user assessments. Findings: Empirical results show that HAD-Finance outperforms traditional rule-based advisory systems and single-agent LLM models. It achieves a 95% recommendation accuracy, an annualized portfolio return of 10.99%, and a Sharpe ratio of 1.25. The system also improves transparency in financial decision-making, contributing to high user satisfaction and trust. Research limitations/implications: While the system demonstrates strong performance, its applicability may vary across different market contexts and investor behaviors. Further investigation is needed into the effects of real-time market data integration, cross-market adaptation, and broader behavioral finance factors. Additionally, user evaluations are limited to a sample of domestic investors in Vietnam, suggesting future studies could expand to international user bases. Originality/value: This study is among the first to combine LLMs with a heterogeneous multi-agent discussion framework in the context of financial advising. It offers an innovative approach to delivering explainable and personalized investment recommendations in emerging markets. The findings contribute to the growing field of AI-powered fintech and suggest potential for integrating HAD-Finance into scalable digital investment platforms, while proposing new research avenues including dynamic personalization and sentiment-aware strategies. ⓒ The Author(s).
Emerging Financial Markets; Factor-Based Financial Analysis; Large Language Models (LLMs); Multi-Agent Systems; Personalized Investment Recommendation