One thing that is maybe underestimated is the very large amount of applied mathematics that is nowhere near the level of frontier mathematics research. For example lots of economics includes a theory component which is usually trivially simple by the standards of mathematics – undergraduate level at best. I tried ChatGPT on the appendix of this paper and it correctly reproduced the results (and the exposition was more elegant):

https://pubmed.ncbi.nlm.nih.gov/35790706/