Terrell Cassada of LoanLogics on AI and Machine Learning
By MBA Insights Staff
January 28, 2019
Terrell Cassada is Chief Information Officer with LoanLogics, Trevose, Pa. He is responsible for document, data and rules processing software development and technologies, as well as corporate infrastructure, data center infrastructure, cloud delivery and information security.
MBA INSIGHTS: What are the differences between AI and machine learning?
TERRELL CASSADA, LOANLOGICS: To the casual observer, there may not seem to be any difference between artificial intelligence and machine learning--many people use the terms interchangeably. However, there are important differences between them, and these differences matter if lenders hope to leverage these technologies to speed loan production, improve the accuracy of loan data and lower costs.
AI describes any technology that applies knowledge to find the best solution to a complex problem. Essentially, AI is capable of making its own choices based on past decisions. In contrast, machine learning describes a system that is trained to complete tasks by leveraging large datasets along with self-learning algorithms and human instruction. But machine learning never actually makes decisions on its own, it simply learns from sample training.
INSIGHTS: Why is it important to define the two?
CASSADA: Both AI and machine learning have enormous benefits in mortgage loan origination, acquisition and servicing decisions, but they are being applied differently. For example, AI is being used to examine datasets and loan file documents to identify the best answer for a credit decision. On the other hand, machine learning is better suited for tasks such as classifying loan file documents and determining the accuracy of data within those documents, which helps lenders maximize performance and improve processing speed.
INSIGHTS: How do AI and machine learning have an impact on the mortgage industry?
CASSADA: I've mentioned a few ways AI and machine learning can be applied to the mortgage process, but there are many more examples. For instance, at LoanLogics, we're currently using machine learning to increase accuracy and automation for data and document processing applications. This helps reduce and even eliminate manual "stare and compare" practices that traditionally accompany loan file indexing and document classification, which frequently lead to loan errors. By using automation to perform these tasks, human operators can focus more of their attention on exceptions. Ultimately, AI and machine learning will also be used in loan underwriting practices, performing complex decisioning that can lead to lower costs and speed mortgage production.
INSIGHTS: How is the mortgage industry different in the way it uses AI compared to other data-intensive industries?
CASSADA: It's true that every industry that deals with large datasets can benefit from AI, from insurance to transportation to the legal industry. However, the mortgage industry is unique from these other industries because of the sheer amount of data lenders and servicers work with on a daily basis, in addition to the many different regulations and guidelines that dictate how they do business. Lenders also have to be very specific when leveraging AI to make credit decisions, especially when they need to comply with multiple regulators in addition to investor guidelines.
INSIGHTS: How does more data volume help in cases of AI and machine learning?
CASSADA: AI and machine learning are useless without data. On the other hand, the more data that is leveraged, the more information AI and machine learning have to do their jobs. The challenge in our industry is that a great deal of data is not that easy to obtain, because so much of it is "trapped" on paper-based forms. While optical character recognition technology can be used to extract data from paper documents, the process is far from perfect, which is why lenders still need human assistance.
Recently, however, machine learning is being adopted to give a hand to OCR technology. Now you have OCR grabbing large blocks of data that are then being fed into machine learning models, which can be trained to recognize patterns in data. From this process, loan data can be more efficiently extracted for validation and verification.
INSIGHTS: What does this mean for the customer experience?
CASSADA: AI and machine learning are both being used to enhance the borrower experience, but in different ways. For example, you see a lot of lenders adopting consumer-facing, digital mortgage products that borrowers can use to try to create a do-it-yourself mortgage experience. While many of these products are great at collecting information from borrowers, they aren't very good at validating and verifying the information along the way. This results in lenders frequently going back to the borrower to ask them for the same information two or maybe even three times, which adds time and frustration to the process.
With AI and machine learning tools, however, lenders can make a quick decision about the accuracy and completeness of a borrower's data as the data is coming in, which results in a faster and more convenient experience for the borrower. On the servicing end, AI and machine learning can also be helpful with determining which borrowers may need help making payments, or for determining when the borrower may be ready to refinance their mortgage or get a home equity loan.
INSIGHTS: Should there be a sense of urgency on the part of lenders and servicers to move forward with AI and machine learning?
CASSADA: Without a doubt. According to the U.S. Treasury Department, AI was one of the three largest areas of investment for financial services companies in 2017. The Financial Industry Regulatory Authority believes AI can help prevent money laundering and help banks provide better data management and customer service. More specific to the mortgage industry, Fannie Mae's most recent Lender Sentiment Survey found that 63 percent of mortgage companies were familiar with AI and machine learning and more than a quarter of all firms reported they were already deploying these technologies.
That being said, relatively few loan QC providers in our industry have dedicated much time and attention to developing AI and machine learning tools. For those that have, these tools have proven to be invaluable. Right now, there is an opportunity for lenders to leverage these new technology innovations, particularly in the areas of data and document processing. Machine learning technology has proven to be significantly effective at classifying loan documents so that data can be extracted and compared to thousands of loan data elements with a lender's LOS data. These lenders are now doubling their auditing speed and reducing loan processing times from hours to minutes. At a time when loan production costs are at record highs, this adds up to real savings. These lenders are enjoying better investor relationships, too, because they are producing higher quality loans. I'd say those are good reasons to get behind these technologies, wouldn't you?
(Views expressed in this article do not necessarily reflect policy of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA Insights welcomes your submissions. Inquiries can be sent to Mike Sorohan, editor, at email@example.com; or Michael Tucker, editorial manager, at firstname.lastname@example.org.)