AI Agents and No-Code Tools in Accounting: A Case Study

Advances in Artificial Intelligence (AI) and Large Language Models (LLMs) have transformed accounting by automating repetitive tasks and enhancing the efficiency of financial reporting. However, their implementation raises challenges related to bias, reliability, and professional adaptation. This article evaluates the comparative performance of three approaches to the vertical analysis of income statements: the traditional manual process, a specialized GPT model, and an AI agent integrating GPT with no-code automation tools. Using the Design Science Research (DSR) methodology, 150 experimental analyses were conducted to measure the execution time, variability, and process scalability. The results indicate that GPT substantially reduced execution time compared to the manual baseline, but still required significant human intervention. The AI agent achieved the greatest gains, reducing the average execution time by nearly 75%, while also demonstrating more stable performance and minimizing the repetitive workload. These findings provide empirical evidence that agent-based automation enhances both efficiency and reliability in accounting workflows, reinforcing its potential to reshape professional practice by reallocating human effort to validation and analytical tasks.