SMARTMONY: UN AGENT FINANCIAR BAZAT PE MODELE LINGVISTICE DE MARI DIMENSIUNI
SMARTMONEY: A FINANCIAL AGENT BASED ON LARGE LANGUAGE MODELS
Author(s): Tudor ZgîmbăuSubject(s): Social Sciences, Economy, Education, Socio-Economic Research
Published by: Fundatia Institutul de Studii Financiare
Keywords: finance; artificial intelligence; chatbot; financial education;
Summary/Abstract: In recent times, we have witnessed an exponential increase in the adoption of tools based on artificial intelligence (AI), accompanied by a growing enthusiasm among experienced developers to launch innovative solutions on the market. Due to its intrinsic flexibility and its ability to mimic human cognitive functions, AI - particularly large language models (LLMs) - has become an ideal choice for a wide range of applications across various fields. Leveraging these advanced capabilities, this paper presents the creation and operation of an AI-powered financial agent. Built on a sophisticated architecture that integrates multiple LLMs interconnected via APIs (Application Programming Interfaces), the system is primarily based on the GPT-4o model but is adaptable to other advanced models as well. Designed to address the persistent issue of financial illiteracy - prevalent across diverse economies - and to support individuals in integrating into today’s financial ecosystem, the proposed agent assists users in completing routine financial tasks, understanding both theoretical and applied financial concepts, and conducting in-depth analyses of specialized documents. Additionally, the system features an innovative "historical" mode that enables comparative analysis of doctrinal perspectives on financial topics. Its development included iterative prompt engineering techniques and both quantitative and qualitative testing with potential users, and its performance was validated through extensive testing across multiple areas of financial expertise. Furthermore, the proposed agent is cost-effective, both in absolute terms and in comparison with many existing market solutions.
Journal: ISF - Working papers
- Issue Year: 5/2025
- Issue No: 5
- Page Range: 1-10
- Page Count: 10
- Language: Romanian
