Anecdotally, I've used Mathematica a bit for prototyping ideas over the years for various computational model prototyping purposes. I rarely see anyone else using Mathematica though, I often too wonder who their typical customer is (outside of educational institutions which its a good fit for).

I find it's great for prototyping certain types of ideas (more abstract/mathematical in nature) because there's a wide range of useful and complex functions that allow you to piece together ideas quickly, iterate/change, and test them. It's quite powerful in that respect.

Sometimes I have an idea and don't want to dig around dozens of libraries, dependency chains, etc. to piece some proof-of-concept test together, nor do I want to attempt to implement everything needed from scratch because that could require a large time investment (and fail conceptually).

Some similar efficient prototyping workflows can now be accomplished in Python using Jupyter/JupyterHUB, at least for me.

Any of my successful prototyped ideas immediately left the Mathematica ecosystem to be implemented using other technology stacks.

Theoretical physics and applied maths. I occasionally see people at the back of seminars with their laptop open, pugging away at their current puzzle with Mathematica.

I've used it to test and prototype calculations and algorithms.

The Stackexchange page for Mathematica is pretty amazing, and quite a good community.

A friend of mine works in theoretical physics and some professor in his groups use Mathematica notation/functions also on whiteboards. I don't know in the field in general though

It's used everywhere chalkboards and differential equations/integrals used to be used.