March 12, 2024

Transforming origination from a human driven process to a software driven one

Mortgage origination, for many years, has meant piles of paperwork, large teams of people performing manual checks, long wait times for consumers, and seemingly endless back and forth. For both borrowers and lenders, the sheer magnitude of resources required often results in a tedious and costly endeavor, with costs to originate a loan sitting in the five figures. Fortunately, some cutting edge lenders have uncovered an opportunity to leverage technology and revolutionize this process.

Mortgage origination is about taking a relatively large dataset (thousands of data points per loan, spanning things like income streams, asset transactions, subject property neighborhood info and square footage) and evaluating a large set of rules against that data (thousands of pages of investor rules and government regulations).

Humans are not good at this

Unfortunately, it turns out that humans, while good at many things, are ill-equipped to internalize thousands of data points across hundreds or thousands of loan applications, and they are also not very good at memorizing and understanding every one of the tens of thousands of rules involved in mortgage origination. While humans bring a wealth of experience, creativity, and interpersonal skills to the table, when it comes to the mechanical side of mortgage origination, there are inevitable shortcomings:

  • Errors: Even the most detail-oriented among us can miss a decimal point, misread a figure, or misplace a document. Especially in mortgage, these errors are extremely expensive when they lead to delays or buybacks.
  • Cost: Humans are not cheap. The cumulative hours spent on “stare and compare” adds up to thousands of dollars per loan. The propensity for humans to make errors drives costs up even farther, because most lenders solve for errors by having multiple human beings do each piece of work to “double check” each other.
  • Scheduling: Humans have limited working hours, necessitating breaks, vacations, and sick days, all of which can delay the mortgage origination process. One problem we hear about often is that no matter how few loans you do, you need at least “two people in each role” because people have to be able to take vacation.
  • Data overload: Humans, as adept as they are, have limitations in processing vast amounts of data swiftly and accurately. Keeping track of all of the incomes, assets, and property details, and knowing all of the guidelines across hundreds of different loan products that can be offered only compounds the difficulties mentioned above, creating higher cost and more mistakes.

Solving those problems with a software orchestrated process

While we're not quite at a stage where every intricate detail can be entrusted solely to computers (especially due to limits in current OCR and document extraction technologies), we can reimagine the process where the orchestration is software-driven. This doesn't eliminate the human touch but rather refines where and how it's applied, focused on things that computers can’t do yet.

In a software-orchestrated process, humans do what the software tells them, usually via a mechanism such as tasks automatically created and assigned by the system. In essence, the administrator is directing their teams not with memos, emails, and standard operating procedures that humans need to commit to memory, but instead with procedures and instructions encoded directly in the systems that users are doing their work in every day.

To take this to the level of an individual processor: At many lenders today, a processor starts work in the morning by looking through a pipeline of loans (possibly for tens of minutes) and reading a bunch of emails in order to figure out which loans are important. Once they’ve honed in on the loan to work on, they need to orient themselves on the scenario and story, and go through the data and documents to figure out which pieces of the loan might be missing, who they need to follow up with, etc. In a software orchestrated process, the processor instead starts by opening a list of tasks, prioritized from top to bottom, and can immediately start reaching out to borrowers or third parties, examining documents that couldn’t be automatically scanned, ordering additional services, or otherwise performing whatever the actual work is.

The benefits of a software-orchestrated approach are multifaceted:

  • Enhanced loan quality: A large percentage of errors in the mortgage process are not “doing the task wrong,” but rather “not doing the task at all.” Missing documents, forgetting to get a verification, or not comparing certain data because the user simply forgot one of the thousands of tasks they need to know by heart is very common. Computers do not “forget” to add tasks to loans, and so this approach dramatically improves the quality of the loans originated. This drives cost savings directly via a reduction in buybacks for lenders.
  • Efficiency: Today, we might have loan officers, processors, underwriters, and closers each checking that the purchase contract matches the LOS for certain data points; that’s the same work being done 4 times, to say nothing of the same user forgetting they already did it at some point over the months-long process and checking it again. Instead of manual redundant processes where tasks are duplicated, a software-driven approach ensures each step is completed exactly once; none are missed, but none need to be repeated. Each task is generated by the system and recorded as done, so it is not assigned to a user again. This means a faster turnaround time and a more streamlined workflow.
  • Seamless change management: Traditionally, any shift in the process requires extensive training and behavior modifications. Operational teams need to be told about the new guidelines, best practices need to be documented and sent out, and then users actually need to read, learn, and implement changes to the way they do work. When users are simply doing what software tells them to do rather than what they are learning to drive via their own memory, administrators and leaders can decide on changes, encode them directly into the system, and see users “following the system” in the new way immediately. This unlocks quick adaptability to regulatory and investor changes, as well as experimentation with different operational processes. This is particularly interesting as lenders might introduce more and more automation over time — if a lender wants to introduce an OCR solution for a new document type, for example, they can simply “assign” that task to the OCR solution in their configuration and no user will ever be prompted to do it again since the automation handles it — there is no longer a need to ask users to “unlearn” their habits.

What's next

There are still many things in the mortgage process where humans are hard to replace — especially when it comes to building relationships and advising the consumer. However, humans today are simply doing too many things that computers are better at — the future of mortgage origination isn't about replacing the human touch, but about optimizing where and how that touch is applied. With the right software paradigms, and particularly with software as the “brain” dictating the work to be done, we can pave the way for a more efficient, cost-effective, and accurate mortgage origination process. Many lenders dream of an end-to-end automated origination, and moving the orchestration of work from a human reading guidelines into the software that already knows them is the largest step towards getting there.