I’ve been dealing with this for years now. You build a solid playwright workflow, it works great for a few months, then the site gets a redesign and suddenly everything breaks. Half your selectors are dead, you’re hunting for new ones, and it feels like you’re constantly putting out fires.
The real issue is that most automation approaches are too rigid. They’re built on fragile CSS selectors or XPath queries that snap the moment someone moves a button or changes a class name. I’ve watched teams spend weeks rebuilding automations after a single UI change.
What changed things for me was thinking about this differently. Instead of hand-coding selectors that get stale, I started exploring ways to generate automations that could adapt. The idea of using AI to craft workflows from plain descriptions started making sense—not because it’s magic, but because AI can understand intent in a way that hardcoded selectors can’t.
The workflows that survive redesigns are the ones built on semantic understanding, not brittle element matching. When you can describe what you actually need the browser to do (“log in with these credentials, then find the download button”), the underlying logic can adjust when the UI shifts.
Have you run into this problem where your automations break after site updates? What’s your current strategy for keeping them maintainable?