Residential Magazine

Uncovering the True Bottleneck

Lenders can accelerate origination tasks with AI processing tools

By Jayendran GS

Streamlining the mortgage underwriting process is among the top priorities for senior-level executives at any lending organization. The current underwriting process is highly manual and has numerous inefficiencies, leading lenders to waste time and money on every loan application.

The actual numbers are even more disconcerting. The average time to process a loan application remains at 45 to 60 days, according to ICE Mortgage Technology. And a recent Mortgage Bankers Association reported that the net loss per loan for independent mortgage banks jumped from $82 per file in second-quarter 2022 to $2,812 per file in Q4 2022.

It is often believed that inefficiencies in underwriting are the primary causes of extended lending timelines, increased operational expenses and a disappointing borrower experience. But it is essential to examine other aspects of the loan origination process to more accurately pinpoint the causes of inefficiencies. Various lenders and underwriters across the industry report that the actual problem lies in the processing stage, which leads to inefficiencies in underwriting.

Processing bottleneck

A high volume of loan applications, coupled with outdated document processing systems, causes a bottleneck. As a result, the loan applications emerging from the processing stage contain inaccurate information, missing data and sometimes even missing documents.

Every underwriter is expected to strike a delicate balance between avoiding the approval of high-risk loans and meeting their monthly performance targets. When inaccurate or incomplete applications end up on an underwriter’s desk, they are left with no choice but to manually go through all the documents to make sure the borrower is eligible for a loan. Underwriters often find themselves having to contact borrowers or originators to obtain the missing information, which is an incredibly frustrating and time-consuming process.

In other words, the output from the processing stage slows down the underwriters, and not the other way around. The ideal solution to this problem is to implement straight-through processing. This would automate the entire lending life cycle, from collection of loan applications and documents to disbursement of the loan, with little to no manual intervention.

But it’s not as easy as it sounds. The mortgage industry must go through various stages of maturity in terms of technology, data purity and process refinement to achieve straight-through processing.

Lending maturity

The maturation of the mortgage industry can be classified into five stages. The first stage is business process outsourcing, where the verification of loan documents is outsourced to third-party companies. The process is resource intensive, slow and prone to errors.

Stage two is logic-based automation. This involves software tools that focus on automating one part of the origination process — the collection of loan applications. It increases the number of loan applications received at the front end but adds a significant amount of stress to other stages of the origination process.

The third stage is the “intelligence layer,” or the use of artificial intelligence (AI) and machine learning technologies to transform the processing layer of the loan origination process and provide contextual insights on a borrower’s creditworthiness. This saves underwriters a lot of time by surfacing critical insights to help them make informed decisions.

The fourth stage is next-generation intelligence, where collected data is automatically cross-verified with multiple third-party services to automatically make informed decisions. The fifth and final stage is straight-through processing, a touchless loan origination system where an application is automatically processed from end to end.

Where we are

Today’s mortgage lenders are largely at stage two. They’ve automated front-office operations to facilitate the touchless collection of loan applications and documents, but they haven’t made much progress in the processing and underwriting stages of the origination process.

There are tools in the market that promise touchless automation without human intervention, but they haven’t accounted for the data problem. Completely touchless automation is not possible until clean data is available to mortgage professionals.

But the mortgage industry can bring significant improvements to the processing stage of the origination process thanks to technologies like AI and machine learning. There are three essential areas where these technologies will play a major role in improving the lending process.

How it could work

Automating the analysis of borrower bank statements with AI and machine learning technologies will help in getting cleanly processed data early in the origination process. The analysis technology should look for suspicious transactions, classify them into various categories and look for discrepancies. By moving these preliminary judgments upstream, the load can be lessened on the processing and underwriting stages.

The analyzed information should still be verified by a human underwriter. But by presenting pre-processed, contextualized information in an easily digestible format, underwriters can be given all the information needed to quickly make a wise lending decision. This can be achieved with the right blend of technology and user experience design, which will make it easy for an underwriter to use the software.

Considering the current level of data purity, it’s certain that companies won’t get 100% pure data from their borrowers. There will always be documents that are not readable or have missing fields. The industry needs an automated workflow that automatically pushes incomplete documents from the AI-powered bank-statement analyzer to a team of experts for human verification.

This will speed up the entire process, as good applications will go to the cross-verification stage after the analysis and only the applications with missing data will be sent for human intervention. Building the aforementioned components into a tech solution will help mortgage lenders shorten processing times, lift the burden off underwriters and improve their productivity while providing the best borrower experience. ●


  • Jayendran GS

    Jayendran (Jay) GS is a co-founder of Prudent AI, a leading fintech platform that is designed to accelerate the mortgage automation process using artificial intelligence (AI). He works with leading nonqualified mortgage lenders to transform their bank-statement loan programs. He is an accountant turned data scientist with rich experience in the finance and technology intersection, including as analytics director at accounting firm EY. He believes in the power of technology, especially AI, to make game-changing transformations to the credit and underwriting processes in the lending business.

You might also like...