The problem isn't the data.
It's the language.
Today's screen is an operator console — it shows the database (steps, journeys, lenses, change-sets) and assumes you already know the model. A newcomer lands on a record list with no path and a wall of jargon.
Every direction below starts from the same fix: say what things are in plain words, and lead with the question the user actually has.
Plain-language reframe
step→Check
journey→Flow
domain→Area
change-set · cs:a8f3→Change history
generation run→AI run
process / component / live lens→removed — filters, not personas
Three jobs the UI must answer
01How healthy is my testing?
02What's broken right now?
03Generate tests for me — the actual value
Three directions follow, each leading with a different one of these jobs. Pick one — or mix. Then I'll build the chosen direction out to full fidelity across every screen.