Playwright UI & Automation Framework

Using AI coding agents deliberately inside the test-authoring workflow.

Playwright · MCP Server · GitHub Copilot

Problem

AI coding agents are genuinely useful for test automation, but only if you know where they help (boilerplate, first-draft locators, debugging flaky failures) and where they don't (test strategy, what actually needs coverage). Most "AI-assisted testing" takes are either hype or dismissal.

Approach

Built a Playwright framework structured so AI-generated test drafts still land in the right architecture instead of sprawling into one-off scripts.
Used an MCP server + GitHub Copilot inside the loop for faster first drafts and faster debugging of failures, with manual review on what actually gets committed.

Results

AI-assisted
first-draft test authoring and failure debugging: Manual
review gate on everything that ships to the suite

Outcome

Faster test creation and maintenance without giving up control over what the suite actually verifies — a workflow, not just a tool swap.

Business Impact

A concrete, honest answer to "how do you actually use AI in testing" — not a hype answer and not a dismissal.