Although the mortgage industry has traditionally been slow to embrace new technology, artificial intelligence (AI) can benefit both lenders and borrowers. There is a great opportunity to improve efficiency, reduce costs and create a better overall lending experience using new advancements in AI technology.
It may seem like a big leap to start using AI in your lending process, but the time to do this has never been better. Originators should know how this technology is changing the industry.
AI — which is computer programming capable of learning (and also sometimes referred to as robots) — is evolving from doing simple, repetitive tasks to being able to perform complex tasks tailored for a specific industry. A prime example of this is the mortgage industry: Cognitive-based AI has a level of industry expertise to enable mortgage processing.
This new generation of AI can make basic decisions and take actions based on best practices. In 2020, the mortgage business will see increasing adoption of this cognitive- based AI. It will be an exciting time of technological advancement in the industry.
For background, the role of a lender’s back office is to perform the required quality and compliance tasks to approve an applicant’s loan. This is true for a credit card, auto loan, student loan, home equity line of credit or a residential mortgage.
For residential mortgages, these quality and compliance tasks are complex and comprehensive because of the size and scale of the loan — a $500,000 mortgage requires a lot more scrutiny than approving a credit card with a $5,000 limit. Also, mortgages are commonly sold by the originating lender to an investor, who retains the right to require that the mortgage be bought back by the lender should the loan end up with quality or compliance defects.
These defects can be discovered by the investor’s own review or later in the process (downstream) should the loan go into default. Examples of defects are missing documents, discrepancies between documents and loan- application data, issues with the home appraisal, etc. In fact, investors that purchase mortgages (primarily Fannie Mae and Freddie Mac) mandate hundreds of defect-elimination business rules that must be executed by banks and mortgage companies.
Today, the vast majority of defect elimination is done manually by back-office loan processors or by business-outsource companies. Staff members have a checklist of several hundred business rules that they work through by collecting applicant documentation, extracting data from those documents, plugging the data into spreadsheets and other systems, and then updating the loan origination system so that a system of record exists to prove (and can demonstrate) that the potential defect was removed.
Once the entire checklist is met, then the loan is ready for underwriting. If you are looking for an answer to the question, “How come it takes 30 to 40 days for my loan to get approved?” the answer is usually this slow, manual process to remove all potential defects.
Automation has made progress in helping back-office loan processors become more efficient, enabling loans to get through their checklist process faster. For example, robotic process automation (RPA) increasingly performs certain back-office tasks for mortgage processors. These tasks are usually extremely repetitive and mundane, and may include data entry, data or information transfers, or the verification of data.
Although automating these tasks was important for improving productivity and reducing costs, they still required humans to step in and make reviews, approvals or other decisions at certain points, because the mortgage lending process is so knowledge driven and requires vast expertise. As an example, there are literally thousands of unique mortgage documents that are used across the U.S.
A process that requires an understanding of what a specific document looks like is well beyond the intelligence of today’s RPA systems. In fact, it’s an intractable problem for current RPA systems. And since most back-office mortgage processes involve mortgage documents, this lack of intelligence and recognition severely limits the scope of today’s RPA systems.
Now, however, through the use of cognitive robots that use deep-learning, neural-network computer vision — and are trained on hundreds of thousands of mortgage documents with specific, embedded industry knowledge — AI is becoming more intelligent to specific industries.
The result is cognitive robots that “understand” the mortgage process, improving efficiency, speed and accuracy while requiring less human involvement. For example, instead of humans executing mortgage-defect rules, robots can now do it either as an “assistant” or as a replacement for a back-office loan processor.
What exactly are cognitive robots? TechTarget defines cognitive robotics as “a field of technology involving robots that can learn from experience, from human teachers, and even on their own, thereby developing the ability to effectively deal with their environment.” In the case of mortgages, they deliver automation at critical points along the loan manufacturing process to provide a productivity boost and improve accuracy.
Because these robots can work on a 24/7 basis without fatigue, they are not only more productive but also less error-prone than humans would be after working long hours and tiring.
There are several benefits to using cognitive robots, which can make a positive impact on your operations and your bottom line. These include learning and recognizing documents, ensuring compliance and increasing the accuracy of handling these documents.
These robots can “learn” and recognize different documents involved in the loan process, such as tax returns, pay stubs, bank statements and gift letters. Once these documents are learned by the robots, they can recognize and process them more quickly the next time they encounter them in a different loan. The robots will know “this is a pay stub from ADP” or “this is a tax return from the state of Arizona.” Consequently, they can process the documents quickly and efficiently. As the robots handle more loans, they get exposed to different documents, so their knowledge base grows over time.
For compliance, robots also learn and gain knowledge of different federal, state and local lending regulations, and can utilize that knowledge to scan for discrepancies, make basic decisions and take actions that help with compliance. Robots can be deployed in a phased approach to allow you to tackle specific pieces of the cycle so that it’s easier for an organization to implement quickly and begin seeing return on investment sooner. Because these robots can work on a 24/7 basis without fatigue, they are not only more productive but also less error-prone than humans would be after working long hours and tiring.
• • •
With cognitive robots, AI can streamline the loan manufacturing process, as well as correspondent loan acquisition and servicing activities, while enabling consistent execution, repeatable outcomes and business intelligence for continuous process improvement. The time is now to embrace AI in your loan processing environment so that you can realize the many benefits that this technology has to offer.