In the 10+ years since the global financial crisis, loads of regulations have been put in place to make mortgages safer for all involved. While mortgages in the US are now much less risky, the cost of mortgage origination has ballooned due to myriad stringent requirements. For mortgage lenders, origination is a long, expensive, document-intensive cycle with many manual processes and repeated steps. The digital mortgage revolution, led by Rocket Mortgage and a growing array of digital point-of-sale (POS) software companies such as Blend and Roostify, has helped improve customer experience through offering self-service apps and online mortgage applications. However, the glossy excitement stops once the buyer documents are submitted. Loan officers, processors, and underwriters then begin their expensive and still overly manual back-office dance of review, verification, validation, hand-offs, and approvals. Lenders need to embrace HFS’ Digital OneOffice framework—linking their digital front-office processes with their back-office origination processes using emerging technologies like automation and artificial intelligence (AI) to drive reinvention.

The biggest pain points in mortgage origination revolve around documents

Mortgage origination, as we know it today, is a complicated set of processes that takes an average of 45 days to complete and costs between $8,500 and $9,000^[1] to process. While loan applications have gone digital and most required buyer documentation is submitted electronically, buyer documents still need to be moved through the origination process lifecycle, which is fraught with complications, including

• Quality. Borrower-submitted documents vary in quality in terms of readability, indexability, etc. Loads of manual intervention is required to determine if the borrower submitted the correct documents and capture tools extracted the right data.
• Validation. Even though the industry has moved from paper-based to digital images and PDFs for loan applications, these files are large (up to 1,000 pages) and complex (about 75 document types). All documents need to be verified to move forward with the underwriter assessment. Ideally, properly trained AI can execute validation and recognize the correct content. But a lack of standard documentation and the quality issues make this challenging.
• Verification. Once the information is confirmed as accurate, third parties need to verify it (e.g., employment verification, asset verification). Aside from being somewhat redundant with validation, there can be issues with automated matching if there is any variability in address or other identifiers.

^[1]Mortgage Bankers Association 2019 

These issues keep lenders from achieving the mortgage equivalent of straight-through processing—“first time right”—where all required information is gathered from the borrower in one go. Exhibit 1 outlines these pain points overlaid on our process map of the mortgage origination life cycle.

Exhibit 1: The US mortgage origination process map, with highlighted pain points

Source: HFS Research 2020

Automation and AI are helping, but we need a greater focus on process reinvention

Mortgage lenders are actively leveraging automation and AI to help optimize the origination value chain. Take the myriad problems with document processing; many lenders are applying emerging tech such as machine learning algorithms trained to recognize specific document types or using RPA to support document verification matching. These task-based approaches yield benefits, but the gains are often incremental and do not necessarily drive process improvement or reinvention. In many cases, lenders follow the same processes, but now with embedded automation or AI elements. This perpetuates process debt—allowing lenders to follow the same old ineffective process. Automation programs should not start with technology. They need to start with a clear set of business objectives and related analysis of whether existing business processes are effective. Technology is only 10% of the challenge. The heavy lifting comes with people and process change.

Business process outsourcing (BPO) firms have been pushing the envelope with process-led change in mortgage origination. Firms such as Sutherland, Infosys, and Indecomm Global Services are leveraging process improvement coupled with automation and AI technologies to reduce the average time to close from 45 days to 30 days. This one-third reduction in processing time drives cost-benefit for lenders and improved CX for borrowers. Some believe the cycle time can shrink as low as 15 days with the right mix of streamlined processes and technology. The path to the Digital OneOffice is paved with process reinvention and enabled by emerging technologies working in concert.

The Bottom Line: Mortgage lenders are making strides toward the Digital OneOffice using emerging tech and process optimization—but they need to think bigger than task-specific change to reap real benefits.

Mortgage lenders have an escalating cost basis and are actively seeking ways to improve their mortgage origination processes. Inefficient document-related processes drive so much of the pain in the mortgage origination lifecycle. Thus, it’s very logical for these document issues to be the focus of process redesign and emerging technologies, which is the real “now” opportunity, albeit one that is very task-specific. If the RPA trend taught enterprises anything, it’s that task automation supports evolution, not revolution. If lenders really want to change, they need a bigger picture view. Here are three thoughts for the future of mortgage origination:

1. A bigger picture. Fixing document processing is progress. But lenders should think about the overall mortgage application and strive for faster decisioning and streamlined validation and verification through consent-based permissions.
2. Make documents irrelevant. If validation and real-time verification grow, then docs become irrelevant, or at least less relevant. If correct information in applications could be confirmed and data could be verified using a social security number or similar, then we could entertain the opportunity to skip processing and go straight to underwriting.
3. Data-based real-time decisions. The future is in corralling the data and leveraging it to make more real-time decisions similar to the credit card industry. We just need to get more comfortable with big-ticket risk management.