I’ve been trying the AI Copilot approach to generate webkit automation workflows from plain English descriptions. The premise is compelling: describe what you want to automate, and the copilot generates a ready-to-run workflow.
I started simple. Described a flow like “log into this site, wait for dynamic content, extract pricing data.” The copilot generated a workflow with navigation steps, waiting logic, and extraction. It actually worked on the first run, which surprised me.
Then I tried something with webkit-specific complexity: “handle Safari rendering delays for lazy-loaded images, retry if selectors timeout.” The copilot generated something reasonable, but it missed some of the webkit-specific retry logic I’d want. I had to adjust it manually.
What I’ve learned is that the copilot generates a solid foundation when you describe what you’re doing clearly. It’s not perfect on edge cases, but it saves massive time compared to building from scratch. The workflow it generates is readable and editable too, so if it misses something, you’re not starting over.
I’m curious though—does the quality of the copilot-generated workflow depend heavily on how precisely you write your description, or does it handle vague prompts reasonably well?