It's not that you can prove that it is exactly the same, which you can't, it's that we have strong and specific reasons to know that it is NOT materially the same, in most situations.

In the area of testing, for instance, I study testers, and I document many things they do which require social competence to achieve. This is tacit knowledge. Tacit knowledge is not automatable. But when you don't know much about testing you don't see any of that tacit knowledge in action. So you think testers are just button pushers whose minds contribute nothing.

When OUTSIDERS to a process claim that they've automated it, they are always wrong; and not just wrong but absurdly wrong.