Recognise spoken word?
I can see possible benefits.
For selected values of possible.
Consider a Glaswegian trying to order from a Portuguese worker who has come to Somerset to learn English. There is enormous scope for confusion.
Now if the voice recognition offered multiple language and dialect options the order could be taken without taxing the language and comprehension skills of the current staff members.
Preparation would then just require a GUI which could be more or less language independent.
I assume the voice recognition system would handle the "but I ordered a large fries" complaints as well.
Still, as Alexa still can't understand a Belfast accent (for example) there may still be significant technical challenges ahead.
Given that the system would need an alternative mechanism to support those with speech difficulties (beyond having a Belfast accent) the alternative mechanism might be the simplest route.