Best AI model for comparing and summarising large numbers of PDF docs?

I have about 350 PDF documents that are condition reports for buildings. The buildings are all of a similar type, and the general content is similar, although they don’t all follow the same structure.

I need to summarise each document by categorising the condition of 5 specific areas in the building – e.g. gutters, roof, etc. These categories will be assigned a rating – good/fair/poor.

The tricky bit is that I want to be able to create a bespoke but consistent rating system, based on comparing all 350 documents against each other, to determine an accurate rating of what is good/fair/poor within the specific scope of these documents. Then I want to be able to automatically apply this to each document – resulting in each of the 350 documents having an accurate summary/rating based on how they fare amongst the other 349 documents.

I feel like this should be something that AI can handle for me, but I’m not sure which tool/platform would be the best to use… any suggestions?