This study investigates the use of Large Language Models (LLMs) in accounting by focusing on their potential to automate financial analysis tasks. The primary goal is to demonstrate how LLMs can streamline accounting processes by reducing the time and complexity involved in analyzing financial data and generating accessible financial reports. This research follows the Design Science Research (DSR) methodology proposed by [1] and involves the development of a Generative Pre-trained Transformer (GPT) designed specifically for performing vertical analysis of the Profit and Loss statement (P&L). The chatbot's performance was benchmarked against traditional methods to assess its efficiency and interpretative improvements. The results show a significant time reduction in the vertical analysis, from 12 min to just 47 s, a 93% decrease. In addition, GPT enhances the accessibility of financial insights for users without advanced accounting expertise, such as managers and small business owners. Limitations include dependency on structured data formats, token and time constraints in large documents, and challenges in user adoption. Despite these challenges, this study illustrates the practical viability of LLMs in accounting contexts and shows the benefits of advanced prompting strategies, including Chain-of-Thought (CoT) reasoning, in improving analysis reliability. This study suggests that LLM-based tools can be extended to other accounting areas, such as horizontal and industry-level comparisons.