A few years ago, the default test automation decision was mostly technical.
Do we use Selenium? Which language? Where do we run the grid?
Today, the menu is much larger. You can write Playwright tests, ask Claude to generate them, use an AI-first testing platform, or combine several of those approaches.
That sounds like progress, and it is. But more choices have also made it easier to optimize for the wrong thing.
Teams compare how quickly the first test can be created. They should be comparing what happens to the hundredth test after six months of product changes.
