1) Semantic search for legal documents. Found it annoying that, as a lawyer, I have to use search engines that require me to guess the keywords that might be used in the documents I'm looking for... as well as to skim through the entire documents the system found in order to judge whether or not they are relevant to my case.

2) Discovering audiobooks I'll enjoy. Someone here on HN mentioned "random sampling" as a great technique to get a feel for a domain without getting biased by one's pre-conceived notions. I thought that would be awesome for books, to literally "not judge them by their cover".

3) Search engine for pre-owned Teslas. Quickly see what's good value based on age/mileage per price, as well as being able to see how long I'll have to wait, on average, to get a particular set of features for a particular price (based on historical market data).

Plus several others.