Mortgage Is Not Being Automated-It’s Being Rebuilt

For more than two decades, the mortgage industry has been in a near-constant state of modernization. Each wave arrived with the same promise: greater efficiency, lower costs, better experiences. And each wave delivered something incremental, but rarely transformative. We digitized documents. We automated tasks. We connected systems that were never designed to work together. Yet despite the investment, the industry today remains burdened by complexity, margin compression, operational friction, and growing strain on loan officers. These are not new observations. They are themes I’ve returned to repeatedly through my work at Transformational Mortgage Solutions and in conversations on Lykken on Lending, particularly in our ongoing Hot Topics discussions. The question worth asking now is not whether technology has failed mortgage lending. It hasn’t. The question is whether we’ve been asking technology to solve the wrong problem.

The Limits of Layered Technology

Most mortgage technology over the past twenty years has been additive. New tools were layered onto legacy workflows, policies, and organizational structures that were built for a different era. The result has been an industry that is more digital, but not necessarily more intelligent. We see this in familiar ways: loan officers navigating multiple systems to complete a single transaction, automation that accelerates tasks without improving decisions, and compliance risk reduced in one area only to surface in another. What makes this especially telling is that the economics tell the same story. Despite decades of technology investment, the average cost to originate a mortgage has more than doubled since the late 2000s, now exceeding $11,000 per loan, according to data from the Mortgage Bankers Association. Digitization did not simplify the system. It often made it heavier. In many Hot Topics conversations, we’ve returned to the same concern: complexity is compounding faster than clarity. Technology alone does not simplify lending if the structure it supports remains unchanged.

Why Artificial Intelligence Changes the Conversation

Artificial intelligence introduces something fundamentally different. Not because it is faster, but because it is capable of continuous learning, pattern recognition, and decision support at scale. For years, AI was discussed as an enhancement. A smarter rule engine. Another tool in the stack. But what we are now beginning to see is a shift away from AI as a feature and toward AI as operating logic. This shift matters because the industry is reaching a human limit. As volumes declined over the past two years, the operational burden increasingly shifted to loan officers themselves. According to industry employment data, loan officer headcount has fallen more than 40% from its 2021 peak, while surveys conducted by firms such as STRATMOR Group consistently show rising burnout and concern about long-term sustainability among those who remain. This is why many of the most thoughtful conversations about AI emphasize not replacement, but support. The future of lending depends on systems that reduce cognitive load, increase confidence, and allow loan officers to focus on judgment, relationships, and guidance rather than constant process management.

The Cost of Partial Transformation

One of the risks the industry faces right now is stopping halfway. Adopting AI tools without redesigning workflows, accountability, and decision structure can actually increase confusion. Organizations may move faster through early stages of the loan process, yet still experience significant rework later. Operational studies have shown that a meaningful share of loans continue to cycle through multiple rounds of conditions and revisions after submission, signaling that speed alone does not equal certainty. In other words, speed without intelligence is not progress. True transformation requires organizations to rethink where decisions are made, how expertise is embedded, and how risk is managed. That is more difficult than adding tools, which is why it has taken so long.

A Case in Point: Building AI-Native Lending Models

In recent conversations with industry operators, including Troy Kennedy, CEO of LoanWorks, Inc., a recurring theme has emerged: the most meaningful change happens when companies design their operating model with intelligence assumed from the beginning. Rather than asking, “How do we add AI to what we do?” the better question becomes, “How would we build this if intelligence were always available?” That mindset leads to different choices: fewer handoffs, clearer accountability, systems that guide rather than chase, and loan officers who spend more time advising and less time managing process. Companies such as LoanWorks, working in conjunction with intelligence platforms like AngelAi, reflect this approach. Not because they use AI, but because they structure around it. The distinction is subtle, but critical.

What This Means for the Industry in 2025

As we look ahead, competitive advantage in mortgage lending will not come from who has the most tools. It will come from who has the clearest systems. Loan officers will gravitate toward environments that reduce friction and increase confidence. Consumers will reward lenders who combine speed with certainty. Regulators will continue to expect consistency and transparency, a trend reflected in supervisory guidance that increasingly focuses on process integrity rather than intent. AI-native models, when designed responsibly, have the potential to address all three. But only if leaders are willing to move beyond incrementalism.

The Question Leaders Must Answer

The question facing the industry is no longer whether artificial intelligence will play a role in mortgage lending. That answer is already clear. The real question is whether organizations are willing to redesign themselves to match what the technology now makes possible. Mortgage is not being automated. It is being rebuilt.