Get a demo
Select the type of demo you'd like to experience with Indico Data.
Select the type of demo you'd like to experience with Indico Data.
Banks, mortgage brokers and others have been processing record numbers of mortgages over the past several years, which would normally be considered a good thing. The bad news is nimble FinTech startups and non-depository loan originators are using automation to process mortgages much faster than their traditional rivals.
Overall customer satisfaction with primary mortgage originators dropped five points (on a 1,000-point scale) in 2021, driven largely by declines in satisfaction with the refinancing process, according to a study by J.D. Power.
To turn it around, banks and mortgage brokers will have to adopt mortgage automation software like those of their FinTech competitors, which use chiefly online tools throughout the mortgage process. Indeed, Rocket Mortgage, a perennial favorite of the likes of Money magazine, is available in all 50 states but offers no in-person service.
Processing mortgages – whether for purchases or refinances – typically involves collecting numerous pieces of documentation from the applicant, taking out relevant data from each, and feeding it into a downstream system, such as a loan origination system (LOS) like Encompass or LendingPad.
The sticking point for most mortgage companies is that extracting data from all those documents is largely a manual process. While robotic process automation (RPA) tools may help to automate some parts of the process, they can only deal with highly structured documents and content. But the majority of documents involved in the mortgage process are unstructured, which means someone has to actually read each one to identify the data to extract.
Intelligent document processing for mortgage underwriting offers an alternative.
A good IDP system is built on artificial intelligence technologies including natural language processing, machine learning and transfer learning. This AI foundation gives an IDP tool the intelligence to read documents involved in the mortgage process just as your employees would. Using models built by those who know the processes best – the employees who actually perform them – the IDP tool can identify and extract relevant data and input it to a downstream LOS.
An intelligent document processing platform can help you greatly increase the speed as well as accuracy of myriad processes involved in mortgage processing, including the following.
One of the first steps in the mortgage process is simply accepting and classifying all the documents that come in from applicants. Typically (but not always) they now arrive via an online portal. But someone must still look the documents over and categorize them according to type, whether it’s property specifications, purchase & sale agreements, W-2s, pay stubs, identification (i.e., passport or driver’s license), real estate bills, insurance forms, and the like. Only then can you send them along to the next appropriate stop for further processing. It’s a process that takes hours for humans but mere minutes for a properly trained intelligent document processing model.
Once all the documents are collected and categorized, the next step is to extract relevant data from them. Here again, this requires an employee to read all the documents to identify pertinent information, then cut and paste it into the downstream LOS or other decision support systems. It’s another time-consuming, rather mundane process. Monotony often leads to errors and certainly to employees who find their work less than rewarding. With IDP, those same employees train models to identify the relevant data to extract and the model handles if from there on, leaving employees free to take on more satisfying and valuable work.
With appropriate data in hand, an IDP platform can automate the process of pre-qualifying applicants by applying their FICO scores. Those with the best scores may be sent on a fast-track route for expedited processing. Fair scores may be sent for further analysis by credit experts while those with poor scores are automatically sent a rejection letter. In this fashion, the bulk of underwriting analysts’ time is spent only on those applications that most require their expertise.
The underwriting process can also be a highly manual one in some institutions, involving preparing and collating documents for analysis, further adding to processing time. This also extends to the closing process, which is often still rife with paper that customers must sign to complete the loan process. Generating all those documents, of course, typically requires manual intervention, making it another candidate for intelligent document processing.
Automation in the mortgage industry can help you break free of these laborious, paper-intensive processes and compete effectively with FinTech’s. To learn more, check out our e-book, “Unlocking the Value of Unstructured Data for Financial Services.”
