I’ve been experimenting with using Latenode’s AI Copilot to generate puppeteer workflows from plain English descriptions, and honestly it’s been a game changer for getting automations up and running fast. But I keep running into this problem: I’ll describe what I want—like “log into this site, navigate to the products page, extract the pricing data”—and the AI generates a working flow in minutes. The issue is brittleness.
After a couple of weeks, the site tweaks their DOM structure or changes a CSS class name, and suddenly my entire workflow breaks. I’m back to debugging selectors and rewriting pieces of the automation.
I know this is a general puppeteer problem, not specific to AI-generated ones, but I’m wondering if there’s a smarter way to handle this. The AI Copilot can generate the initial workflow, sure, but how do you actually make it resilient to those kinds of layout changes without ending up maintaining scripts constantly?
Does anyone have a pattern for this? Like, are there better selector strategies, or ways to structure the workflow so it’s more adaptable?