Mortgage Application Income Verification Demo

We are going to talk about the income verification stage and why it is right for automation-driven improvements. Income verification is a critical step in the mortgage origination process and is heavily reliant on a human’s expertise and decision-making abilities. This process typically involves a review of a mortgage application, multiple recent bank statements, pay stubs, and tax returns to validate consistent income, employment, and the ability to repay. It could also involve additional documentation such as board letters, employment verification letters, pensions, trusts, or other tax returns.

When a human is involved in this process, they likely use multiple systems such as email, Excel spreadsheets, loan origination systems, and file folders. They must copy and paste, classify each document, and extract the data to ensure all requirements are met. This involves cross-referencing the data between these documents and manually transferring data into another system.

At Hyperscience, we have converted this traditionally manual and repetitive process into a digital experience. Hyperscience engages the human only for tasks where the solution is not yet sure what it needs to do, if it has rules in place to defer to a human, or when it feels the machine learning could benefit from more human involvement to assist with further training. Hyperscience will classify and separate incoming documents to ensure income verification standards are met, notifying a human only where necessary.

Let’s see how this task can be simplified with the assistance of Hyperscience from two different personas: the potential borrower and the mortgage processor. First, as the individual applying for a loan, my loan officer asked me to complete a loan application, which I did here. This is handwritten in blue pen. Besides the application, I also submitted tax return documents, pay stubs, W-2s, and a bank statement via email or a secure drop folder so the mortgage company could decide on the loan.

As the loan processor, I know these documents have arrived and Hyperscience has started working on them. I can assist Hyperscience where necessary. Hyperscience digitizes all the data elements provided and runs specific business tools to ensure this information is within all compliance and risk thresholds, giving me direct guidance. Hyperscience has classified the files by document type and pulled the data. In most cases, it has done this without human help, but there is one case where the machine is not yet certain how to process the document, so it flags it for me.

In this case, Hyperscience asks me to review one of the bank statements. The machine knows almost everything about bank statements, but it is calling my attention to a specific area. It is not sure if this is something it should include. It has captured a line meant to be just a total. As a human, I confirm that this should not be included as a withdrawal or deposit. Everything else the machine was able to get, and the color coding helps me navigate this quickly.

That was the only thing the machine needed help with. Now it is going to apply the defined business logic. It ensures we have all documents and that the dates align with the application. We need two months of bank statements, two months of pay stubs, and the last two years of tax returns. Hyperscience goes through all the data points and suggests decisions. In this case, the mortgage originator has decided that a human needs to be the decision-maker, perhaps for regulatory reasons, the audit trail, or the company’s risk profile.

The flow continues and applies this business logic. It asks specific questions: Do we have those two previous years’ W-2s? Does the stated income between all documents vary less than 5%? It also verifies that we have the last two months of bank statements and the pay stubs. As the human operator, I can see the four decisions I am waiting to make. The machine makes a recommendation in each case. It confirms the W-2 dates are valid. It calculated the stated income variance was less than 5%, perhaps accounting for a mid-month pay raise. It also confirmed the correct bank statements and pay stubs are present. My only action is to review the documents and see the suggestions from the machine, allowing me to quickly reach an actionable status.

I am going to complete this task. When I do, it sends all this data downstream to the mortgage origination system or archival system. The income validation has been approved because I selected valid for all four rules. Beyond just verifying the income, Hyperscience extracted all the handwritten data on the other forms and normalized it to a useful state. This data now goes to the downstream system, reducing the risk of clerical errors and the time associated with that data.

To recap, Hyperscience has automated this process and completed all tasks within the income verification workflow with speed and accuracy, guiding the human towards critical items. Beyond the basics of income verification, Hyperscience also digitized the loan application and moved it directly into the downstream systems to better drive towards a straight-through origination process.