AI Data Analyst Agent for Indian SMEs

Abstract

Small and medium enterprises (SMEs) and educational institutions in India generate large volumes of tabular data, such as sales records, GST and Tally exports, academic results and placement statistics, yet most of them lack the expertise needed to analyze it. Existing chat-with-your-data tools are generic and are not designed for the inconsistent formatting of real Indian business data. This paper proposes an AI Data Analyst Agent, a web-based agentic system in which a user uploads a tabular dataset and asks questions in natural language. The system profiles the dataset and uses a multi-agent architecture, consisting of a Query Planner, a Code Generator, an Insight Explainer and a Chart Recommender, to plan the analysis, generate code and explain the result. AI-generated code runs inside an isolated Docker sandbox, while dedicated machine learning models, namely Prophet, Isolation Forest and K-Means, handle forecasting, anomaly detection and clustering. Results are returned as plain-language explanations with automatically selected visualizations. Because the intelligence lies in the analysis process rather than in dataset-specific logic, the same agent can serve sales, academic and institutional data without redevelopment. The proposed system aims to make computation-backed, explainable data analysis accessible to non-technical users through an open-source technology stack.

Keywords