Sundeep Mathur: Mortgage banks must use AI “compliantly and responsibly”

PERSON OF THE WEEK: As the mortgage lending industry rushes to embrace advanced AI, there are principles that should be followed to ensure a long-term, sustainable roadmap for success.

AI can help lenders increase revenue, improve operational efficiency, and ensure regulatory compliance – but the technology and tools are evolving at a dramatic pace.

To learn more about how AI is changing mortgage lending, MortgageOrb recently interviewed Sundeep Mathur, Vice President of Fintech Business at TAVANT.

Q: How should lenders sort through the different opportunities and areas for applying AI and set priorities? Where should the North Star be?

Mathur: Advanced AI is transforming industries by leveraging vast human knowledge and enabling natural language interactions. Unlike previous tools, today’s AI solutions are revolutionizing the way the industry works, offering tremendous opportunities to increase revenue, reduce costs, and ensure regulatory compliance. In credit and banking, the potential applications are diverse – from improving marketing strategies to automating operations that currently rely heavily on human intelligence.

However, given the variety of AI applications available, banks and lenders must choose carefully what they focus on. As key players bridging Wall Street and Main Street, our primary goal should be borrower empowerment. This means using AI not just for operational gains, but also to inform and empower borrowers. By bringing them closer to the business of lending and giving them confidence in their financial decisions, lenders are fostering more informed and active borrowers. A parallel can be drawn to the securities industry, where citizen brokers have significantly expanded market participation.

Ultimately, the guiding principle in applying AI should be an unwavering focus on improving the borrower experience. Prioritizing AI solutions that empower and support borrowers will not only lead to their success, but will also drive growth for lenders and the industry as a whole. By making borrower empowerment the guiding light, lenders can ensure that AI integration leads to long-term prosperity for both our clients and their institutions.

Q: What controls and measures do lenders have in place to ensure their AI solutions meet regulatory compliance requirements?

Mathur: Lending has always been a highly regulated industry, with rules designed to ensure the safety and soundness of borrowers, lenders and the industry as a whole. As technology advances, particularly in the field of AI, these regulations become even more important. While AI offers tremendous potential, it also brings challenges, particularly when used by malicious actors or when unintended consequences occur.

From the beginning, the focus on compliant and responsible AI has been a rallying cry for our society and industry. We are witnessing a concerted effort by academia, the AI ​​industry, and government entities to ensure that AI is developed and deployed responsibly. Regulatory attention at the state and federal levels is intense, and policies are being developed to address critical issues such as explainability, bias and discrimination, opt-out options, and alternatives to human involvement.

Academically, the focus is strong: over three million students are studying AI, including over three thousand at the doctoral level, and are dedicated to advancing safe AI practices. At the same time, the industry is investing heavily—between $100 billion and $150 billion annually—in expanding AI capabilities, building stronger base models, and improving controls and expertise.

These joint efforts have resulted in solutions such as RAG (Retrieval-Augmented Generation) and vector databases that help anonymize protected information and secure personal data. Another result is the RAG Assessment Framework, which allows us to evaluate AI models for accuracy, relevance, and security. Such advances ensure that AI-driven solutions are implemented responsibly, safely, and compliantly.

Q: How can lenders identify AI initiatives that deliver ROI?

Mathur: In lending, there is a clear link between IT spending and business growth. Traditionally, investing in technology has been the engine of growth, but today IT spending is essential for survival. The introduction of artificial intelligence (AI) marks a critical shift. Lenders that do not actively pursue AI solutions risk being overtaken as competitors gain a significant lead through advanced AI capabilities.

While the cost of AI will eventually come down, lenders need to recognize its value now. Waiting is not an option. AI can transform key business processes and create opportunities that are too valuable to ignore.

To fully realize the potential of AI, it is important to identify where it can provide the greatest benefit. There are four key areas where AI excels:

Analyze and summarize: AI can sift through massive amounts of unstructured data, provide insights, and answer questions that would otherwise require extensive manual analysis.

Predicting the future: By analyzing past data, AI can create models that predict future trends and enable lenders to make better, data-driven decisions.

Automate actions: AI can bring intelligence to automation, allowing systems to make decisions and initiate actions, reducing the need for human intervention.

Content creation: AI can generate new content from patterns in text, images and other data to support marketing, customer loyalty and product development.

Lenders need to identify where AI can deliver the greatest ROI in their business areas. From marketing to servicing, AI is already delivering significant benefits. The potential is enormous and now is the time to act. By starting today, lenders can ensure they realize the full value of AI and remain competitive.

Q: ChatGPT seems very powerful. Why can’t a lender just use ChatGPT directly in their business?

Mathur: ChatGPT and similar large language models (LLMs) are undoubtedly revolutionary tools. These models leverage vast amounts of internet-based knowledge, enabling them to understand natural language, engage in logical reasoning, and provide users with a conversational experience. The results are often impressive, with AI providing quick, seemingly simple answers to complex questions. However, this simplicity can be deceptive.

Although ChatGPT can provide answers to virtually any question, it is important to remember that it does “its best” to generate an answer. The model places emphasis on providing an answer rather than ensuring absolute precision. In addition, the content shared with it becomes part of its learning process. For companies, especially in the credit and banking sector, the use of these tools requires a more rigorous approach.

The stakes are high in the mortgage industry. Lenders must comply with strict regulatory requirements, manage risk effectively, and ensure AI is used responsibly. Securing proprietary policies, pricing, and customer data such as PII is critical. In addition, providing consistent, accurate analytics and responses is paramount. Continuous monitoring is essential to identify and address any biases.

While ChatGPT is incredibly powerful out of the box, it is not always perfectly tailored to lenders’ needs. There is a chance that it may lead to anomalies or behaviors that require mitigation, so expertise in leading tools and technologies is essential.

Using ChatGPT effectively is like owning a new smartphone – it’s only really useful if it’s personalized. Adapting and fine-tuning these models to meet specific business goals is key to unlocking their full potential.