Measure twice, cut once: this adage is apt in your situation. You stand to benefit hugely from AI implementation but you have so much data success will be dependent on system design.
Some questions are discrete and supervised learning based algorithms would be apt. For example 'Properties likely to loaned against'. The second question in such a scenario is likely to be 'How much would the lender extend?".
This is based on property value and therefore it a continuous value, in which case unsupervised learning would apply. Factor in 'Effect of litigation on value of a property' and alpha-beta tree pruning becomes a prudent option. Then there is the question of what is the best infrastructure developing, testing and deploying the AI. What about consolidating and sanitizing the dataset(s)?
Of course
My point is that we need to first establish your requirements, detail them then figure out the best algorithms to implement. And if that is the case then a the best person for the job. Am an experienced programmer, software architect and a computer science graduate.
Please get in touch to discuss this very interesting project in detail.