The rapid increase in the volume, variety, and velocity of data, especially in the automotive industry, has created a strong demand for easy-to-use analysis tools. These tools aim to enable non-experts to gain insights without depending entirely on data specialists. However, the complexity and heterogeneity of data make this challenging, particularly when dealing with both structured and unstructured sources. Recent advancements in Large Language Models (LLMs) have enabled natural language interfaces for structured databases. Still, current solutions often fall short in evaluation mechanisms, feedback transparency, and integration of unstructured data such as images and Light Detection and Ranging (LiDAR). This thesis addresses these limitations by proposing a novel, multi-agent architecture, Analyzer Chat Bot, designed to facilitate seamless interaction with both structured and unstructured data sources. The system incorporates specialized agents for data access, workflow routing, evaluation, error handling, analysis, and output formatting. It supports feedback transparency, interpretable query generation, and enhanced usability for time-series and visual data. Through this architecture, we explore how LLMs can access, integrate, and summarize information across data types to directly answer user queries. Our contributions include a new evaluation pipeline for multi-agent LLM systems, improved prompt handling, integration of metadata explanations for structured data, and visual output for unstructured data. The results demonstrate an adaptable and user-friendly approach to bridging the gap between data complexity and non-expert accessibility.