For example, check whether there is any difference in the scrap rate between the two factories.
If you feel different and want to verify it, then the zero hypothesis is "no difference", and the confidence that it is "different" after rejecting the original hypothesis is 95% (if α=0.05).
For example, to test whether the new drug is better than the old one.
We must choose "no difference" as the zero hypothesis, because we must be very cautious when approving new drugs, and we can't make the conclusion that "the new one is better than the old one" without great confidence. Choosing the zero hypothesis of "no difference" is actually protecting this conservative hypothesis (zero hypothesis and opposite hypothesis are not equal).