Data wrangling is a critical yet time-consuming phase in data science, often consuming up to 70–80% of total effort. Existing tools require significant technical expertise and lack intelligent automation. This paper presents Axel AI, an agentic AI-based system that automates the complete data wrangling pipeline using natural language interaction. The system integrates Large Language Models (Google Gemini) with a multi-agent architecture consisting of intent understanding, transformation planning, safe execution, and validation modules. It separates LLM reasoning from deterministic Python execution to ensure accuracy, reproducibility, and safety. Additionally, the system supports Excel intelligence features such as formula generation, chart creation, and dashboard building. Experimental results demonstrate significant improvements, including a 75% reduction in processing time and enhanced data quality. Axel AI enables both technical and non-technical users to transform raw datasets into analysis-ready formats efficiently, making it a powerful solution for modern data-driven applications.
Uplanchiwar et al. (Thu,) studied this question.