Laying Foundations for the Digital Mortgage of the Future

By Rajesh Bhat, Co-Founder & CEO, Roostify

While most of our recent focus has been on the increased digital adoption during the pandemic, subtle changes in the mortgage industry over the past decade have resulted in a process today that is unrecognizable from even a few years ago. The ability and ease of home ownership has undergone a seismic shift that is still underway. It’s the underway part where our industry still gets a lot of blowback, and for good reason. 

While groceries and goods can be delivered with a few clicks of a button in less than an hour and a ride can be summoned with an app, the ability to obtain a mortgage or refinance an existing one, while a lot better than it used to be, remains a cumbersome and un-automated process for millions of Americans and their lenders. 

Truth be told, a completely automated, real-time solution that takes a mortgage from application to approval to disbursement without the touch of any human hands in operations is still a long way away. Just as Uber and Amazon.com require real people to oversee technology, automation, and even navigation, the mortgage industry still needs people to oversee, assess, review, and confirm the information required to help people secure the funding they need for the biggest investment of their lives - and to do so safely, securely and without risk of fraud or predatory lending practices.

At Roostify, we steadfastly believe in designing technology that helps people realize the dream of homeownership, and will continue to do so. As we look towards a truly digital mortgage in the future, our goal is to create features that get lenders and borrowers excited and foundational, innovative solutions that make everyone’s lives easier.

Our newest product, Roostify Document Intelligence (RDI) Service, the first service on our new Roostify API platform, enables home lenders to instantly classify, validate, and extract data from mortgage-related documents utilizing artificial intelligence. RDI is currently able to process more than 25 of the most common borrower-provided documents (e.g., paystubs, asset statements, tax forms), and by the end of the year, we will have upwards of 50 mortgage document types covered. 

Digitizing data from documents replaces the need for manual data entry, a repetitive, time-consuming, and costly action that creates operational inefficiencies and opportunity for errors in the mortgage underwriting process. More importantly, it creates the opportunity to understand the “time value” of the data. In other words, being able to access, process, compare, and augment this data early on (days and, sometimes, weeks earlier) eliminates steps, expedites the loan eligibility process, and delivers more pristine data to the loan origination system.

Historically, the extraction of data from digital documents has almost always required manual review to attain a confidence level of 90% or more on the data extracted. The focus of our partnership with Google Cloud (GCP) in building out the Document Intelligence Service was leveraging GCP’s Lending Doc AI capability to achieve a similar confidence level without manual work.

Across the entire document set, the average confidence level is  greater than 90%, with 35+ document types having a confidence level of over 95%, without manual review. Thanks to machine learning capabilities, this confidence level only improves over time (approaching 100%) as we process more documentation and correlate extracted data to loan application data. That speed and certainty of the data extraction makes real-time validation possible and data-driven automation within reach.  

The Document Intelligence Service is readily available to all lenders as well as mortgage service providers and technology providers. It can support an infinite number of use cases benefitting from automated identification and extraction of data contained within the mortgage document set. It then compares and validates this data against the mortgage application data set.

We’ve been using the Document Intelligence Service with some of our own clients and have integrated it into our Lending Platform at the beginning of the borrower's experience. By doing this, we dramatically expand the loan data set early in the lending journey through document data extraction and create true data-driven task decisioning based on patterns and rules. You can learn more about it here.

Document processing may not be the shiniest, or most exciting piece of technology imaginable for improving the mortgage process. However, as we look towards a truly digital mortgage in the future, the operational efficiencies this kind of technology enables serve as a foundation, and an absolutely fundamental one, for future innovations. Indeed, looking into the next decade, innovations in AI, machine learning, and other technologies will lead to a home lending industry that is virtually unrecognizable from today’s, and the foundations that we lay now are just the start of that process.