A product UI that never sits still
HeyGen's product surface is enormous and changes constantly: the script-to-video editor, avatar and voice selection, generation queues, preview rendering, team workspaces. Features ship in hours, and UI components are regenerated by coding agents as fast as the models behind them improve.
For HeyGen's QA team, that pace had a price. Playwright suites bound to specific selectors broke on every refactor: tests that passed on Monday failed on Tuesday because a button got renamed. Authoring and repairing that suite consumed 60% of the Head of QA's time, and manual verification of what the suite missed consumed much of the rest.
From selector upkeep to intent-based YAML
HeyGen replaced selector-bound scripts with Shiplight's intent-based YAML tests. A step in a HeyGen test describes what the user is doing, not which CSS selector to find: open the editor, paste a script, click generate, verify the preview renders.
When the UI underneath changes, Shiplight re-reads the page and re-resolves the intent instead of failing on a stale locator. An intended redesign heals silently; a real regression still fails loudly. The suite survives exactly the churn that used to destroy it.
The same agents that build HeyGen now test it
HeyGen's engineers build with AI coding agents, so its testing layer works the same way. Through Shiplight's MCP server and Skills, the agent that ships a change verifies it in a real browser, writes the YAML test for it, and repairs the test when it drifts.
Coverage stopped being a separate project owned by one overloaded QA lead. It grows as a byproduct of shipping, authored by the same agents that write the features.
Maintenance all but gone, time back for hard problems
The result is the number in the headline: time spent authoring and maintaining Playwright tests fell from 60% to almost nothing within a month. That time now goes to the technical work a Head of QA should be doing: hardening release quality rather than babysitting selectors.
For AI-native teams like HeyGen, that is the real promise of AI-native testing: quality infrastructure that moves at the same speed as the product it protects.
“I used to spend 60% of my time authoring and maintaining playwright tests for our entire web application. I spent 0% of the time doing that in the past month. I'm able to spend more time on other impactful/more technical work. Awesome work!”
Jeffery King
Head of QA, HeyGen