From my research and experiments, I think that ML languages help a lot "getting the specs right". That means that you have more confidence that the code does what you want it to do.

I can see a lot of value for that in non-tech companies that correctness is crucial, like finance, insurance, health...