everything that uses predictions? it correlates with mass ofc so you need a certain amount of these problems that predictions make sense.
So maybe Phone companies incoming requests for cancelling your contract.
you could also go for network optimizations. Complex model calculations -> what is needed that a town in country XYZ can optimize it's water consumation.
ML is just statistics so think about something where you have enough information to train such a network, or there is enough money behind it so it pays off in the long term to invest money into it.
That would be my approach.