I work within the field of NLP and data logistics, but most of the information is stored in the description. I mostly use NLP algorithms to extract the sentiment or the similarity if they have a similarity score calculation. For example, something like the windshield cracking, or something extremely severe like a garage burning, or multiple car damage. 

The NLP tasks don't deal with the financial data directly but use extra information to get some insights. It's not really a good practice in India to use deep learning models directly on financial data because of audit reasons.

It is mostly used for everything that is connected to machine learning, data science, training, and other similar things.