Taking Control of the Data Supply Chain

The quality of data that initiates loan origination processes is fundamental to everything that follows: operational productivity, time to close, risk in the lending decision, meeting the demands of the secondary market, profit margins. 

 And yet, compared with carmakers, mortgage lenders don’t enjoy anything approaching the level of control over their supply chain. Instead, lenders take whatever documents are provided and run quality checks. They must extract data about the borrower, assess the veracity of the documents, and verify the accuracy of the data extraction process itself.  

Mortgage operations professionals know that conversion of documents into data and data verification are laborious, complex yet mission-critical tasks. The result is a data bottleneck as the need for quality is inevitably restricted by human capacity and accuracy. In addition to creating risk, inaccurate data raises costs further along the production line. When accuracy is off by just 5 percent, productivity decreases by 30 percent or more.  

The data bottleneck has long frustrated operations leaders, but solutions have been hard to find. Expanding in-house teams was traditionally the best option for adding capacity without compromising quality, but this loads the business with highly salaried employees – not ideal at a time of volatile demand.   

Outsourcing offers flexible capacity, but quality is harder to manage and labor costs are rarely reduced. Offshoring used to be a lower-cost option, but fees have risen as other economies have matured. Meanwhile, there’s increased concern about the risks of transferring customer data overseas. 

Specialized Technology for Skilled Tasks

Returning to our manufacturing analogy, carmakers have successfully used automation —robotics — to boost productivity when human skills are unavailable or too costly. Many lenders have also invested in automation platforms with a goal of achieving touchless automation. But the docs-to-data and verification components of these systems have disappointed with poor accuracy. Human specialists are routinely needed to step in and figure things out. 

Constant human monitoring of automation technology undermines its core value. Operations leaders are left to figure out some awkward combination of people and technology in the hopes of increasing productivity or else give up on automation altogether.   

Some lenders have discovered a different option: TRUE. 

TRUE provides task-focused artificial intelligence for docs-to-data conversion and verification. Like skilled human operatives, the TRUE technology is trained in the lifecycle and language of loans. It is uniquely able to review documents contextually, seeing the connections between the borrower and their data, and building a complete picture.   

This allows TRUE products for docs-to-data conversion and data verification tasks to slot neatly into data supply chain processes and deliver fully automated performance, with no humans in the loop.  

Task specialization means TRUE products consistently surpass human performance. The AI models have been trained on lending industry data, so they deliver at least 90 percent accuracy out of the box

The AI improves on this foundation, learning the specific needs of different lenders and loan manufacturing processes, so accuracy levels exceed 99 percent within a matter of months.  

Soon enough, operations leaders see that they can trust the accuracy of automated data, allowing them to redeploy skilled human resources to where they can add more value. 

 With TRUE, lenders can efficiently scale up in busy times without the burden of high fixed costs in fallow periods. The result is a mortgage-making factory that’s in control of its data supply chain, ensuring consistent quality and readily adapting to demand.