Aditi Subbarao, Global Financial Services Lead at INSTABASE, at AI for Financial Services NYC – GenAI adoption in loan origination

Aditi Subara, leader of financial services at Instabase, uses a pop quiz to contrast Michael Faraday's scientific contributions with Elon Musk's impactful technology application. This highlights the goal of turning AI's potential into tangible business value for financial services, akin to Musk's impact with Tesla.

Instabase and AI's foundation

Instabase helps organizations convert unstructured data into actionable insights, addressing the issue of scattered data. AI, like an electric motor, relies on data. Most enterprise data is unstructured. Early AI in financial services focused on process optimization, automating tasks previously done by humans. Initially, AI extracted data from documents, but generative AI now integrates and analyzes data from multiple documents. This is exemplified in loan origination, where AI tallies data from pay slips, bank statements, and employment letters to determine a borrower's income.

Expanding AI's potential

AI's potential goes beyond document processing to creating comprehensive views by integrating both internal and external data. This next phase aims for automated decision-making with safeguards. Organizations have vast data sources and numerous potential AI applications, focusing on high-impact areas where most time is spent on data processing. A framework with three pillars is outlined: operational data processing, specific data querying, and knowledge search, aiding in systematically identifying AI applications across organizational functions.

Operational data processing

Automating data extraction and structuring from various documents improves efficiency in processes like loan origination. AI can create chatbots for querying static datasets, such as underwriting guidelines or expense policies, showing it's flexibility. Advances like ChatGPT enable AI to interact with data, answer complex queries, provide deeper insights, enhancing data analysis. Effective AI solutions require integrating language models with visual and layout awareness to handle complex enterprise data comprehensively.

Ensuring rigorous testing and governance

These are crucial to ensuring AI accuracy and reliability. AI systems should be vetted and equipped with confidence scores and human review mechanisms. Not all tasks need sophisticated AI solutions, so organizations should decide where to apply AI based on benefits and risks, avoiding over-reliance on large language models. Applying AI involves identifying appropriate use cases, systematically applying AI across operational, querying, and knowledge functions, and ensuring thorough testing and governance.

AI in loan origination

AI's capabilities in loan origination include automated data extraction from complex documents, transforming them into structured data for downstream systems like loan origination systems, pricing engines, or risk management systems. An example from Instabase during the COVID-19 pandemic illustrates this: an American bank faced a surge in loan applications through the PPP lending program and used Instabase to build a solution in six days, increasing processing capacity from 10,000 loans a day to 10,000 loans an hour, showcasing AI's potential to significantly uplift productivity in operational processes.

Specific data querying

Building dedicated chatbots for specific functions within an organization, such as querying underwriting guidelines. Chatbots handle fixed data queries, providing precise responses. With conversational AI users can interact with data, asking questions about specific documents or datasets and receiving detailed, contextual answers.

Governance and risk management

A robust governance framework, data security policies, and rigorous testing protocols are essential to validate AI outputs. Organizations must ensure AI models are accurate, transparent, and accountable, capable of explaining their workings and tracing data sources and calculations.

Strategic application of AI

Strategic application of AI involves careful selection of use cases, applying AI across different organizational functions, ensuring thorough testing and governance. Focusing on areas with the highest business impact, leveraging AI for operational efficiency, specific data querying,and integrating robust governance frameworks,organizations can harness AI's full potential.