Design and Implementation of a Telegram-Based Conversational AI Agent for Automated Personal Expense Management Using Large Language Models and Google Sheets

Abstract

Manual logging of day-to-day personal expenses is often abandoned by users because it requires them to open a dedicated spreadsheet or application and enter data in a rigid, structured format. This paper presents the design and implementation of a conversational Artificial Intelligence (AI) agent that allows a user to record and retrieve personal expenses simply by chatting in natural language through the Telegram messaging platform. The system is built using the n8n low-code automation platform and combines a Telegram Trigger node, an AI Agent powered by an OpenAI Chat Model, a Simple Memory module for conversational context, and two Google Sheets tools, namely Read Expenses and Add Expense, that allow the agent to write new expense records to, and retrieve existing records from, a connected Google Sheet. When a user sends a message such as "I spent Rs.231 for Medicine today," the agent extracts the expense description, amount, and date, appends the entry to the spreadsheet, and returns a natural-language confirmation. The agent can also generate an on-demand categorized daily summary. The proposed system was implemented and tested with real conversational data, and the results demonstrate that the agent accurately parses free-form expense statements, updates the backend spreadsheet correctly, and produces coherent, human-readable summaries, thereby lowering the friction traditionally associated with manual expense tracking.