Tag: AI evaluation

When it comes to mortgages, accurate data is non-negotiable. It’s simply impossible to maintain regulatory compliance or confidently sell loans on the secondary market without it.  

However, ensuring data accuracy can be a daunting task. It takes a tremendous amount of time and effort to review loan documents and check and cross-check data manually. At the same time, the traditional technologies lenders and servicers have used to identify and capture data from loan documents are extremely limited.  

Adding to this data dilemma are two recent market developments that have made it even harder for lenders and servicers to ensure data accuracy. Originators are facing new QC requirements from Fannie Mae, while the hot MSR market and the growing potential for increased defaults is placing pressure on servicers to get data right. But as the French playwright Molière once noted, “The greater the obstacle, the more glory in overcoming it.”  

 

The Latest Push for Loan Quality 

Quality data demands on lenders rose significantly this year when Fannie Mae introduced new pre-funding and post-closing QC requirements that are expected to go into effect next month. The GSEs’ new policies require lenders to complete a minimum number of pre-funding QC reviews every month, plus lenders must now complete their post-closing QC reviews in 90 days instead of 120 days. 

While there has been some pushback on Fannie’s new requirements, they underscore the critical need lenders have for tools that enable them to quickly identify incorrect or missing data that could lead to loan defects and potentially costly repurchases. It also bears mentioning that Fannie Mae is far from alone in demanding higher loan quality. Investors and federal regulators are increasingly requiring it, too.  

Invariably, this means lenders need accurate data extraction technology that enables them to instantly identify, capture and extract data from any loan document. The problem? Most third-party providers of data extraction technology and legacy optical character recognition (OCR) tools have failed to keep up with the growing number of structured and unstructured loan documents. Some providers have been cutting staff in response to higher rates and lower loan volumes, too, so the problem continues to persist.  

 

Servicers Have Their Own Challenges  

The need for accurate data is not limited to lenders. On the servicing side of the business, high quality data is equally paramount—and growing more so. 

One reason for this has been the growing demand for MSRs, as many organizations have pivoted to servicing loans to counter decreased loan production. Ensuring accurate data is critical to quickly ingesting and analyzing loan files and auditing loan documents before integrating them into a servicer’s system. 

Data accuracy will grow significantly more important if mortgage defaults rise, too. Such a development is looking increasingly likely, since many new homeowners have found themselves stretched thin financially due to the recent rise in interest rates and inflation. In order to perform loss mitigation processes efficiently, including forbearance requests or loan modifications, servicers will need to ensure all existing loan data and the data from incoming borrower documents is accurate.  

Unfortunately, many data extraction technologies have proven no more effective for servicers than they have for originators—either because they are built on older technology or because their solutions are unproven. And ensuring data accuracy when exchanging MSRs or determining the proper loss mitigation option for a borrower in default is equally as intimidating when performed manually as it is for originators. Not only do manual processes create delays and potential errors when onboarding loans, but they also add to a servicer’s costs in what is already a cost-intensive business.   

 

Harnessing the Power of AI 

With the proper technology, any mortgage organization can ensure it is working with accurate data and is able to seamlessly identify potential issues before they become expensive problems. The right technology can also help lenders identify mismatched data and use it to help spot fraud. 

The key lies in leveraging modern, proven AI-powered data extraction technology that enables companies to effortlessly classify and extract data from diverse structured and unstructured documents with speed and precision. That is why an increasing number of lenders and servicers are turning to Paradatec’s award-winning AI-Cloud solution. 

Equipped with an impressive pre-built library of over 8,500 unique data fields found in loan documents, AI-Cloud harnesses advanced machine learning capabilities to swiftly and accurately transform the information trapped within documents into individual data elements. And it does so extremely accurately and reliably, so that clients can generate purified data to fuel their automation initiatives and gain valuable insights into their processes and strategic plans.  

Because it can pull data from over 850 known mortgage document types within seconds, AI-Cloud also empowers clients to confidently audit large volumes of loan files to meet Fannie Mae’s new requirements and post-closing deadlines. It also enables servicers to automatically audit data provided by prior servicers against incoming documents before ingesting MSRs into their system, thus minimizing errors and enhancing their overall customer experience. 

 

If recent market developments have you concerned about data accuracy, we can help. Just drop us a note at info@paradatec.com! 

As the housing market takes a breather, most mortgage and real estate companies are turning their attention to reducing costs. And few processes are more costly or time-consuming than document management.

When it comes to improving efficiency, shortening the time between receiving loan documents and making use of them stands at the top of the list. Fortunately, there are many document management software options for streamlining operations, including products that leverage optical character recognition (OCR), AI and machine learning technologies. Yet there are just as many broad and bold claims being made about them. How do you know what to believe? Or better yet, how do you test these technologies?

To help, we list three important questions to ask when evaluating these claims. Before you begin comparing vendors though, you’ll need to ask yourself a few questions.

 

Know Thyself

Mortgages may be a commodity of sorts, but every company is unique in terms of their business goals and how they get things done.

Before looking for document analysis technology, first ask yourself, “What specific problems are we trying to solve?” To answer this question, you must know what your requirements are on a high level. What workflows and documents are you using? What data needs to be extracted? And how do you intend to use that data? Which systems will be involved?

You’ll also need to know what your decision factors are. For example, how much are you prepared to spend on implementation and ongoing cost of ownership? How quickly do you plan to implement the solution? Is there an upcoming deadline for an audit you need to prepare for?

Once you know what your needs are, you can evaluate vendors by asking the following three questions:

1. What will I actually get?

Simple question, right? Not exactly. If you’ve done any research into document technology vendors, you’ve encountered mountains of happy marketing lingo and hyperbole. What you need are facts, so this question demands a few follow up questions.

For example, will they do a live demo with you? Will they use documents you provide? What results will you get, and how are they measured? How granular are those results, how confident can you be with them, and what tools help you evaluate them?

Keep in mind that the mortgage and real estate markets aren’t static. What works today may not work as well tomorrow. So, what is the vendor’s approach to confirming results on an ongoing basis? The minute you hear, “you can set it once and forget it,” it’s probably time to look elsewhere.

2. How does the system improve itself?

Now we’re getting into the nitty-gritty. The key here is understanding if and how results are improved over time, which is what you really want. You’ll also want to know how the accuracy of the results are measured and whether those measuring tools are available to you.

If you find the vendor uses machine learning, deep learning, transfer learning or convolutional neural networks to constantly improve results, you may be on the right track. If the vendor is using these technologies in a layered approach, that’s even better. But then you should ask whether the vendor built these technologies or licenses them. If they’re licensed, the vendor may not be able to explain how the system improves itself.

3. What can I do with the results?

This is the time to find out whether your needs will be simply met or surpassed. For example, how easily are the results available for you to use with other systems, such as your system of record, data visualization software or reporting tools? What can you do to improve the results? How easy is it to implement changes to increase the number of documents recognized or the amount of data that can be extracted?

Finally, you’ll want to know how proactively the vendor adapts to change. After all, new document types and data needs are constantly evolving with new lending requirements, investor guidelines and market demands. How often is the technology updated? How frequently does the vendor make new releases available? How are they preparing for anticipated market changes? Are they willing to spend the time with you in a true partnership?

 

Gauging Future ROI

If you do this exercise thoroughly, you’ll have all the information you need to make a smart decision. Ideally, you‘ll be able to gauge future ROI in the form of quantifiable and qualitative savings.

Quantifiable savings is determining the amount of time in human process hours you’ll save by leveraging the solution, as well as how much you will shorten process cycles by reassigning staff to other areas. Qualitative savings can’t be measured as easily, but are equally valuable. It’s what you gain through outcomes such as enhancing risk management, using extracted data to feed data analytics initiatives and improving the customer experience.

At the end of the day, finding document technology that fits and is “real” is not an easy task, but it’s an essential one. At Paradatec, we’re always ready to answer your questions about our AI-based document technologies—plus we have real case studies showing they work. To learn more, please contact us at info@paradatec.com.