The problem
Today's tools draw the arrow from the wrong end
Every team that automates end-to-end testing ends up with the same four pains. None of them is about the framework.
They are the authority — and have no time to keep an inventory of features and scenarios.
A moved button breaks the only written record of the requirement.
Days after the developer moved on — the dependency was knowable at the merge request.
Cheaper tests, same blind spot: nobody knows what to test.
A real case
At a company we work with, users who registered with email + password landed on "Sign-in cancelled". The real state was "email not verified". Nobody had written the scenario registered → not yet verified → allowed in, so there was no test and no screen for it. The product lead found it by asking an AI to walk through the flow and record it — a scenario tree, drawn by hand, once, after the fact.
When you find out
The break is visible at the merge request. Today it is found at the release.
Sound familiar?
Tell us which of the four you have. That conversation is how we pick the first teams.