MBA Premier Member Editorial–The Real ROI of Digital Mortgage: Putting Loan Officer Capacity Where it Matters

Skaidre Trakimas is CEO of Pragma USA

Over the past several years, the mortgage industry has poured real money into digital front doors: application portals, e-signature flows, document upload tools, borrower dashboards that promise a modern intake experience. Yet for many lenders, the return on that investment is measured in the wrong currency. Applications started, site visits, portal logins are activity metrics. The more important question is whether digital intake improves pull-through, lowers the cost per funded loan, and allows expensive human capacity to be deployed where it can actually change the outcome.

That gap matters more now than it did a few years ago. According to the MBA and STRATMOR Peer Group Roundtables Program, independent mortgage companies spent an average of $12,209 to originate a single loan through the retail channel in 2025, down 15 percent from the 2023 peak, but still a number that makes every inefficiency expensive.

Sixty percent of that cost sits in sales expense alone. That makes loan officer capacity one of the most expensive resources in the origination model. The strategic question, then, is not whether technology should replace the loan officer. It is where human judgment creates enough value to justify the cost of deploying it. In other words, the most costly part of origination isn’t the paperwork behind the scenes, but everything that happens between a borrower’s first click and a funded loan. That is exactly where the digital front door lives, and where lenders are most likely to get the automation-versus-relationship question wrong.

The Front Door Is Not the Operating Model

Most lenders already have a digital front door. The harder question is whether it has actually changed the operating model behind it. Too often, institutions digitize the point of entry while preserving manual handoffs, duplicate reviews, and downstream exception paths. The result is a better-looking borrower experience layered on essentially the same cost structure.

The stronger model treats the front door and the loan officer as parts of the same system. Technology should absorb the administrative weight of origination, collecting identity and income documentation, routing straightforward files through automated checks, giving borrowers real-time visibility into where their application stands so that loan officers spend their time on the conversations that actually require judgment: structuring a loan for a self-employed borrower, explaining a rate lock decision, working through a title issue.

When technology carries the predictable work, it does not diminish the role of the loan officer. It concentrates that role on the moments where judgment, advice, and intervention can materially affect the outcome.

How Does Hybrid Onboarding Keep Mortgage Intake Personal?

Frictionless doesn’t mean fully self-service. Effective hybrid onboarding pairs automated intake with clear signals for when a human needs to step in and routes the borrower to that person before they abandon the application, not after.

Consider a self-employed borrower or an investor with multiple business entities. That borrower is statistically far more likely to stall while entering income documentation or disclosing ownership structures than a borrower with a single W-2. A well-designed system detects that hesitation and alerts the assigned loan officer in real time, with visibility into what has already been submitted and what remains outstanding rather than forcing the borrower to start a support ticket from scratch or, worse, simply leave.

That suggests a different way to design the borrower journey: around complexity and intervention value. Straightforward files should require minimal human involvement, while files with identifiable friction should trigger targeted intervention. More complex borrowers or structures should receive high-touch support when human judgment can materially improve the outcome. Treating every borrower journey the same is expensive for the lender and often frustrating for the borrower.

This only works if the institution defines the rules in advance: which events trigger human assistance, who responds, how quickly, what information they already have, and which situations require additional review beyond a first response. Done well, this is straight-through processing with a human touch: automation carries the predictable work within a file, while people are deployed precisely where complexity, judgment, or exception handling actually requires them. The distinction matters: automating tasks is not the same as automating loans.

Digitizing Intake Does Not Automatically Reduce the Cost to Originate

Complex intake doesn’t just frustrate borrowers; it adds up in cost. A smooth-looking application form is not enough if the information it collects doesn’t flow directly into underwriting, compliance, and servicing systems. That is the difference between digitizing a process and redesigning it. A portal enhances borrower experience without affecting origination economics if the same info is repeatedly interpreted, validated, re-entered, or reconciled during the loan lifecycle.

The KPIs that matter here are not the ones most dashboards default to. Application completion and conversion still matter, but they are insufficient on their own. Leadership should also be looking at time to initial decision, manual touches per file, document rework rates, exception-resolution time, pull-through, and how much capacity is freed up per loan officer.

More importantly, those metrics should be understood by borrower journey, product complexity, and acquisition channel rather than only as enterprise averages. Given that sales expense already accounts for 60 percent of the cost to originate at independent mortgage companies, and fulfillment another 20 percent, understanding where capacity is being consumed is not just an operational question. It is directly tied to unit economics and cost per funded loan.

A lender that knows its average cost to originate but cannot identify which borrower journeys are consuming disproportionate capacity is missing the management insight needed to improve it. Depositories have a different cost structure, but the management question is the same. Sales expense runs closer to 42 percent, corporate and production support another 38 percent, and fulfillment the remaining 20 percent.

Those proportions matter because they show that inefficiency does not sit in one function alone. If technology improves the borrower interface but the same information is still reviewed, reconciled, or re-entered across sales, production support, and fulfillment, the institution has digitized the experience without materially changing the underlying cost structure.  

A condition cleared once in a portal and re-verified twice more downstream isn’t a minor inefficiency; multiplied across thousands of files a year, it’s the difference between a lender that improves its cost to originate year over year and one that doesn’t.

Digital Acquisition, Measured the Right Way

It’s tempting to treat a rising number of digital applications as evidence of a healthy acquisition strategy. It isn’t, on its own. A lender can grow digitally initiated applications while pull-through stays flat, manual intervention stays high, and time to clear-to-close doesn’t move. Digital adoption is not the objective. Economically productive digital adoption is. A new borrower relationship creates value only once it’s active, operationally sustainable, and aligned with the institution’s credit and risk appetite, not the moment a form gets submitted.

This is where regional and mid-size lenders have a real, underused advantage. Competing with national digital originators does not require matching their self-service model feature for feature. The advantage comes from being deliberate about where to compete on automation and where to compete on expertise, while still delivering the speed and transparency borrowers now expect as a baseline—reserving high-touch service for the moments (jumbo loans, non-QM structures, complex self-employment income) where it visibly changes the outcome. High-touch service creates the most value when it’s applied with intention, not when it’s spread evenly across every file regardless of complexity.

Automation Changes Decision Rights, Not Just Workflow

A digital front door changes more than the borrower’s interface. It changes how information enters the institution, how work moves between sales, operations, underwriting, compliance, and risk, and how decisions get documented along the way. That makes governance a design question, not a post-launch cleanup task.

Before automating a single workflow, leadership needs answers to a short list of questions:

• Who owns the end-to-end borrower journey?
• Which tasks can be automated outright, and which require human review or escalation?
• How are exceptions and overrides documented, and by whom? Which team is accountable when response times slip?
• And when a file falls outside the rule, who owns the exception?

The quality of an automated operating model is not tested on the perfect file. It is tested when income does not reconcile, documentation conflicts, eligibility changes, or an exception crosses functional boundaries. Without clear answers, each function optimizes its own piece of the process (sales for speed, compliance for coverage, throughput operations).

In contrast, no one owns the borrower’s experience or the loan’s ultimate performance. Automation can execute a rule perfectly and still produce a bad outcome if the rule, the data behind it, or the escalation path never had a clear owner.

This is also where governance stops being an internal efficiency question and becomes a compliance one. Every automated decision point in the borrower journey (an auto-approved condition, a system-generated disclosure, a routed exception) needs to be as auditable as a decision a human underwriter would have made manually.

From a risk and compliance perspective, the quality of the borrower interface matters far less than the institution’s ability to reconstruct what happened, why a decision or action occurred, what data supported it, and who or what was accountable. For leadership, that means four things need to be explicit from the start: decision rights, exception ownership, data lineage, and auditability.

The Way Forward

The lenders best positioned for the next phase of the market will decide,  journey by journey, where automation should remove work, where human intervention changes the outcome, and where judgment itself is part of the product. That model must evolve as products, borrower segments, and regulatory requirements change. With MBA projecting single-family origination volume to approach $2.2 trillion in 2026, the opportunity is to handle more volume without allowing cost and operational complexity to grow at the same rate.

For leadership teams, the key question is clear: Which borrowers can proceed with little supervision? What events require human input? And does human intervention genuinely enhance outcomes like pull-through rates, cycle times, risk management, or financial results?

If a lender cannot answer those questions with data, it may have digital tools, but it has not fully redesigned digital origination. The real value of technology is eliminating unnecessary work and concentrating people where their judgment creates the most value.


(Views expressed in this article do not necessarily reflect policies of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes submissions from member firms. Inquiries can be sent to Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)