Transforming loan trading with generative AI

When considering loan sales, unstructured data can come in the form of pdf files or Excel sheets, loan contracts, collateral valuation reports, court communications or any other document typically found in a virtual data room of a complex loan sales transaction. At NPL Markets, we have started testing the use of AI on a data mapping exercise of non-performing loan data tapes to the new data template published by the European Banking Authority for NPL sales transactions. We extracted key value pairs from court documents and collateral valuation reports and enriched loan data tapes with details about the geographical vicinity of real estate security.?

However, the flaws of existing LLM are well known. AI generated answers may not be accurate or outright wrong and the LLM training data may not include the most recent market developments. It is still uncertain how secure it could be sharing sensitive client data with open AI platforms like OpenAI’s ChatGPT or Google’s Bard. However, the market is evolving at an incredible speed. Deploying an LLM in a private cloud using a dedicated cloud and AI infrastructure from providers like Amazon Web Services could help overcome privacy concerns. Bloomberg recently announced an LLM that focuses on finance and claims to be able to analyse financial disclosures or to conduct sentiment analysis and credit risk assessments.?

Once these crucial issues will be fully investigated, financial institutions might leverage AI-enabled fintech platforms to improve their management of credit portfolios and loan sales through the platform’s ability to efficiently process and analyse vast amounts of unstructured data.

At NPL Markets, we are fully committed to analysing the use of generative AI and serving our international customer base with better transaction advice, data enrichment services and valuations while strictly safeguarding our client data.