[2A1-GS-10] AI application:

[2A1-GS-10-05] Automation of End-to-End Test Processes Using LLM in CI/CD Pipeline

LLM-Driven Automation for End-to-End Testing in Conversational Systems: A Case Study

〇Shigeki Nagaya¹ (1. Technology Development Division, Neural-Group Inc.)

Keywords:LLM, Software Engineering, E2E Testing, AI Agent

This study aims to address the challenges of automating end-to-end (E2E) tests based on user scenarios in CI/CD pipeline. To achieve this, we propose a system architecture that leverages large language models (LLMs) to fully automate the testing process—from unit tests (UT) to E2E tests. Our approach extracts test design details from test specifications using LLMs, enabling the automatic generation of both test code and reports. Integrating this system into the CI/CD pipeline significantly reduces test preparation and execution times while ensuring rigorous quality assurance. Experiments on an in-house product confirm that the proposed method enhances test generation accuracy and reduces operational costs compared to conventional practices. These results indicate that our approach offers a promising solution for comprehensive test automation in agile development environments.