Lenders should rethink their priorities for digital mortgages, use AI

0
79

Sharp Credit – Credit News – Credit Information

For years, mortgage lenders have struggled under the burden of two, traditionally conflicting directives: lowering loan costs and improving loan quality. Quality is pretty much at an all-time high, as lenders have complied with steadily increasing loan quality requirements. But profitability has become a major challenge. In what has become a “new normal,” many mortgage CEOs now tell me their primary concern is how to make money.

With origination volumes fluctuating and margins shrinking, lowering costs is a necessity. And yet, to ensure consistent loan quality, many lenders have backstopped their aging manufacturing technology with human experts and processors to manually check and recheck loan files and documents. Not only does this increase loan production costs, it leaves lenders open to human error and a lack of loan quality transparency.

In other words, the “new normal” costs too much. Human-enabled technology practices have created a double hit on the expense line and record high origination costs. We may feel more confident about avoiding fines, fees and penalties by removing loan defects — but the cost is clearly not sustainable.

New fintech solutions haven’t really helped, either. In recent years, lenders have invested mightily in new consumer-facing technologies designed to fill up pipelines, improve the consumer facing experience and shorten time to close. While “speed sells,” so to speak, lenders are finding the real problem lies with what lurks beyond the point of sale.

Under the covers is a myriad of complex tasks that rely heavily on the collection, interrogation, processing and use of the enormous amount of data that drives all decisions in the mortgage lending process. This data must be accurate and consistent in order to drive effective automation. And yet, the manual way in which data is being checked and rechecked does not make this task very easy.

Just over the past two years, however, the financial services industry has seen the development of new tools that have everybody talking and have provided a ray of hope: artificial intelligence, or AI, and machine learning. Together, these tools have already proven to dramatically impact a mortgage lender’s operational efficiency and the accuracy of information.

AI covers a range of technologies and analyzes data to identify patterns and make choices based on previous decisions. As a subset of AI, machine learning leverages large datasets and self-learning algorithms to complete tasks with a high degree of accuracy. Both tools promise to transform every stage of the mortgage lifecycle. In fact, machine learning is already eliminating the manual tasks of classifying loan documents and determining data quality. The results have been beyond comparison when it comes to processing data — regardless of format — then structuring, validating and verifying that data.

For example, machine learning has given new life to OCR (optical character recognition) technology. OCR alone has been limited in its capacity to classify and extract data. This has forced lenders to resort to stare-and-compare tests to determine the veracity of loan files and build templates that rely on such things as “keywords and anchors” to properly identify documents and the location of data to extract. When fortified with machine learning tools, however, OCR is used to broadly capture data so that machine learning can use it to classify exponentially more documents and allow more data to be extracted with far greater accuracy.

This purified data can then be used to ignite rules-based automation that enables compliance auditors to increase the number of daily audits they can complete from around six to 20 or more. There is no reason why this approach cannot also be applied to underwriting, the last and most difficult milestone in our industry’s pursuit of the digital mortgage. This ultimately will improve a lender’s digital mortgage offering for consumers and will help to dramatically lower production costs.

According to a study by PwC, by 2017 more than half of financial services firms had already invested in AI and machine learning. But according to Fannie Mae’s Mortgage Lenders Sentiment Survey released in October, just more than a quarter of mortgage lenders had deployed these tools. Our industry can do much better.

It’s time to face the moment of truth. The mortgage industry needs AI, machine learning and other digital innovations more than ever, and these tools must be fed by quality data. The traditional means of collecting and evaluating data, while ensuring loan quality, are no longer sustainable, and the “new normal” is no longer acceptable, either. By adopting these new tools, our industry may finally be able to turn the corner on its loan quality issues and skyrocketing costs for good.

Want to learn more about digital mortgages?


For reprint and licensing requests for this article, click here.




Original Source