I wonder if it's more about convincing yourself that it faithfully follows the same workflow an analyst would follow. It's always possible to miss stuff, so the

I wonder if it's more about convincing yourself that it faithfully follows the same workflow an analyst would follow. It's always possible to miss stuff, so the best a person or a machine can do is be demonstrably methodical, it sounds like... and that is easier to test. Unless there is really some magic tacit step that human analyst perform to get better answers.