A Unified Platform for Resolving Citizens’ Queries on Beneficiary Services by Using AI‑Powered Chatbots

Abstract

Limited access to government beneficiary schemes exists for rural citizens because they have limited access to clear information and digital literacy skills. This investigation demonstrates a conversational AI system which processes natural queries from citizens before understanding their objectives and obtaining qualification elements and recording unhandled concerns. A MultiLM sentence encoder operates in combination with logistic regression for intent classification while DistilBERT acts as a token classifier and rule-based functions verify eligibility against existing laws. A microservice architecture using FastAPI with MongoDB processed 107653 CPGRAMS records while reaching a 0.92 macro-F1 for intent detection together with 0.89 eligibility precision and a 60 % faster response time than manual procedures. The open-source pipeline contains synonym augmentation to tackle class imbalance and reproducible scripts and anonymized data for enabling large-scale affordable implementation. The obtained results prove that fast and trustworthy public-sector chatbots can create transparent digital governance systems which operate efficiently and build trust among citizens.