Supply chain
One of the first ways that machine learning was demonstrated effective, was when machine learning was proven to be more effective at identifying malignant melanomas than top medical experts.
An interesting article on identifying problems in a supply chain takes this further. Machine learning could be trained in identifying and faulty products in a supply chain, making quality control much more efficient.
10 Ways Machine Learning Is Revolutionizing Supply Chain Management
Crime
There are too many CCTV security cameras for human eyes to actively monitor. Footage is recorded and then played back later if an incident is reported.
Machine learning technology, having being trained with past footage, will enable machine learning to detect where there is abnormal or erroneous behaviour than what is normal for that particular environment. The computer algorithm would then draw attention to what is going on, making the cameras a tool for disrupting a crime incident taking place, rather than just being an evidence gathering tool after an incident has taken place.
To extrapolate this to the nth degree, pre-crime may be possible by detecting suspicious behaviour. Just as machine learning currently can detect melanomas better than human experts, machine learning in the future may detect changes in environments that seemingly foresee crimes about to happen.