When you make a substantial technology investment for your company, are you thinking: “I’d like to deploy a system where I can take the most mundane aspects of my processes and get better at them?” Or do you think “I’d like to eliminate the most mundane aspects of my process.” At first glance, it may seem like a semantical spin on the same concept, but in practice the two statements lead to very different approaches and outcomes.
Consider the process of on-boarding documents for loan origination and underwriting. For years lenders have turned to legacy document capture vendors who addressed this business challenge with increasingly complex and sophisticated programs. From document digitization (“Scanning, no need for filing cabinets!) to OCR (“No more manual data entry!”) to RPA (“robots will automate document classification and data extraction!”)
What each of these evolutionary steps in technology have in common (and any technology cycle that takes over two decades to develop is at best evolutionary) is that they address the first iteration of the fintech investment quandary: they are applications that make it faster and easier to convert documents into actionable data. In exchange for your purchase and investment in time to learn and implement the software, you will automate a major portion (usually over 50%, if the website claims are to be believed) of the human “work” involved in a process.
Read the case studies posted on capture vendors websites. They usually explain how they shortened loan approval times, allowing customers to close more loans with less work, or how they use to have 20 data entry operators and now they only need 10 to complete the same number of loans. It’s great to be more productive, but essentially you are still in the data entry business.
This is why legacy capture software often acts as a tool and not a solution. To point out the differences between tools and solutions:
- a solution allows you to focus on the intellectual, critical core aspects of your business and remove supporting tasks from consideration;
- a solution doesn’t require you or your staff to be trained on its usage; and,
- a solution doesn’t take months to implement.
Of course, there is value in investing in tools. But to get to the point where you are creating the “eliminating mundane aspects” rather than “getting better at mundane tasks” you need to evaluate technology investments as a solution.
As long as processes are initiated by documents, there will be a requirement to review documents and extract data. But when it comes to a tool versus solution approach, when onboarding borrowers for example, you should look to remove data entry from the process.
Instead of the tool approach:
- Scan documents
- Perform OCR
- Review and make corrections
- Connect to Your LOS
The solution approach is:
- Capture documents in your LOS
- Data is populated in your LOS
Of course, it isn’t as simple as scan and convert, but as a platform, or more appropriately as a service call from the LOS, it should appear to be a black box solution. When building loan origination solutions, your philosophy should center on several fundamentals:
- Develop a deep understanding of the critical document types required for loan underwriting.
- Use the latest machine learning and AI technology to classify documents and extract relevant data.
- Incorporate data validation experts within the process to direct robotic processing for data extraction.
- Offer this service as an API call within existing core mortgage systems.
In practice humans are not removed from the data capture process, they are integrated into it, using attended machine learning (AML). The difference is that humans are part of the process, and specialize in data analytics, rather than data entry operators, or even worse, loan underwriters. The AML model fulfills the promise of capture solutions: eliminating manual data entry, allowing lenders to lend.
