Multilingual Conversational AI for Financial Assistance: Bridging Language Barriers in Indian FinTech
Bharatdeep Hazarika, Arya Suneesh, Prasanna Devadiga, Pawan Kumar Rajpoot, Anshuman B Suresh, Ahmed Ifthaquar Hussain
TL;DR
The paper tackles the barrier of linguistic diversity in India's fintech sector by designing a multilingual conversational AI capable of Hinglish code-mixing. It proposes a novel multi-agent architecture with a Language Classifier, Orchestrator, specialized financial tools, and a Response Generation Module that decouples language from core financial logic via query rephrasing. Through an empirical evaluation of language models and a real-world deployment, the work identifies Indic-BERT for language detection and Hermes-3-8B for response generation as strong choices, achieving multilingual task success on par with English. A proof-of-concept deployment with 500 beta users demonstrates significant gains in task completion, session length, and retention, illustrating the practical impact of vernacular digital financial services in India.
Abstract
India's linguistic diversity presents both opportunities and challenges for fintech platforms. While the country has 31 major languages and over 100 minor ones, only 10\% of the population understands English, creating barriers to financial inclusion. We present a multilingual conversational AI system for a financial assistance use case that supports code-mixed languages like Hinglish, enabling natural interactions for India's diverse user base. Our system employs a multi-agent architecture with language classification, function management, and multilingual response generation. Through comparative analysis of multiple language models and real-world deployment, we demonstrate significant improvements in user engagement while maintaining low latency overhead (4-8\%). This work contributes to bridging the language gap in digital financial services for emerging markets.