Musinsa Reduces Regression Test Time from 1 Day to 30 Minutes Using Playwright and AI-Based TC Generator

The Musinsa Retail Platform QA team resolved regression testing bottlenecks during regular deployments by building Playwright, a Slack bot, and an AI-based test case (TC) generator. Previously, manual testing took a full day, but tests are now completed within 30 minutes within the automated scope.

Playwright-Based UI Automation Strategy

Playwright was chosen over Selenium and Cypress due to its auto-wait, network interception, and Trace Viewer features. Specifically, the Page Object Model (POM) pattern was applied to lower maintenance costs when UI changes occur, and Okta authentication sessions were shared per worker to optimize execution time. Smoke tests and regression tests were separated to quickly verify core payment flows, while edge cases are intensively verified before regular deployments.

AI-Based Test Case Generation and Slack Integration

A pipeline was built where inputting planning documents allows AI to visualize test structures via mind maps and generate specific test cases after review. Claude is utilized for interpreting complex domain logic, while Google Gemini is used for generating repetitive patterns, with both models run in parallel to increase speed. Additionally, a Slack bot @doomto was developed so that team members without CI/CD access can trigger tests via commands and share results in real time.

Results and Future Plans

Currently, 41 files and 234 test cases have been automated, covering payment and member-type scenarios for complex domains such as mPOS. The time required to write test plans and TC documents has also been significantly reduced. In the future, the team plans to expand functionality beyond the Retail Platform to the entire PBO domain.