I think you're pretty on point here; today human nuance is needed to decide WHAT to test as well as HOW MUCH to test. It's also useful for executing some types of tests too (which is why we support both automation, and human testing). IMHO Unit and integration tests should be a given. Human testers should be used with the highest leverage where possible. Being a functional-tester though, today is partly limited by what tooling is accessible to you - which we want to fix.