Perhaps this is one of the understanding gaps that crop up around AI development? At my current company and most others I've worked at, testing capability is part of the same bucket because engineers do their own QA.

I'm far more interested in understanding how we can 10x our confidence in a change and not just our line counts.