NewsLink Q&A With TRUE’s Ari Gross: How AI Prevents Borrowers Data Errors, Reveals their Root Cause

Ari Gross

As CEO and Co-Founder of TRUE, Dr. Ari Gross’ vision is to leverage AI, data and automated processes to transform the entire lending experience. With over 20 years’ experience in artificial intelligence, computer vision, and data science, Gross drives the TRUE team daily to reimagine how technology can change the entire lending lifecycle, leading to smarter and quicker decisions, and previously unimaginable experiences for lenders and borrows alike.

Prior to TRUE, Ari led CVISION Technologies, where he focused on image compression, OCR and business process automation. He then launched SoftWorks AI to address the growing needs of the knowledge worker automation market, which evolved into TRUE with a focus on lending intelligence.

MBA Newslink: Your business seems highly focused on borrower’s documents. Why is that?

Ari Gross: Ultimately, the benefit we bring to the market is faster and better lending decisions. However, the challenge for lenders is obtaining the data they need from within borrowers’ documents. Business processes that use highly accurate digital data can become fast, reliable and efficient.

These outcomes translate into better products and better margins, so every financial services business has maximized its use of digital data and carefully manages data quality. That includes many forms of lending: personal loans, auto loans and credit cards. Mortgage lending is the exception because the data in borrowers’ documents remains physical and requires manual processing. This severely limits mortgage lenders’ ability to originate loans using fast, reliable and efficient automation.

MBA Newslink: What’s your view on mortgage lending being a digital data laggard?

Ari Gross: The reason is simple: mortgage lenders process the most complex data sources of any financial services business. Borrowers’ documents are the polar opposite of digital data. In some cases they receive fully digital PDF documents, but more commonly they get scans or photos of physical documents. It’s only quite recently that we’ve had machines capable of reading images of documents with reasonably accuracy, so lenders have had no choice but to train people to read the documents and enter the relevant data into loan origination systems.

MBA Newslink: Many lenders and their customers like the human touch. Doesn’t automation take that away?

Ari Gross: Where you apply people in a process is often a matter of choice. The complexity of borrowers’ documents had prevented that being an option. That means less resources for high-value work like underwriting or customer service, which are among the reasons why home loans take a long time to close. When borrowers’ documents can be understood automatically, lenders can apply human skills to tasks that are more valuable or appreciated.

MBA Newslink: Can machines really match human skill in processing borrowers’ documents?

Ari Gross: The short answer is yes, but overpromising by some technology vendors has led vendors to be skeptical. That’s why TRUE is so focused on borrowers’ documents – our lending intelligence solution was built for this category of data and no other. Our focus, along with our heritage is artificial intelligence for understanding documents, means our accuracy rates are 95 percent or more at the application stage, rising higher as we check and recheck data through the loan manufacturing process. We also improve our performance over time by adjusting our models to meet the needs of individual lenders.

MBA Newslink: What changes when borrowers’ data becomes more accurate?

Ari Gross: Data quality really is the key to faster and better lending decisions. When data is complete, accurate and readily available, the entire lending process improves, from origination to closing and final quality control for the secondary market.

The longer data errors persist through, the more the cost-to-correct grows. This is a source of cost that comes right off the bottom line. Worse still, errors can result in buyback which are even more costly.

MBA Newslink: Is there any other intelligence gained from lending intelligence?              

Ari Gross: Our clients find insights in our data that goes beyond our core purpose. Some are using our technology to reveal the weak spots in their processes, pinpointing when, where and the exact nature of data errors. Operations leaders can understand if there is a technical fault, a training issue or some other hidden factor. It’s especially valuable for getting to the root cause, especially for errors that are recurring but intermittent. This combination of prevention and cure means lenders can use our AI to forensically improve quality as well as reduce costs.

(Views expressed in this article do not necessarily reflect policies of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes your submissions. Inquiries can be sent to Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)