The mortgage industry is facing structural pressure that incremental improvements alone cannot resolve. Mortgage origination costs have risen sharply over the past decade, margins are razor-thin and rate volatility is punishing fixed-cost operating models. At the same time, lenders are under immense pressure to modernize mortgage operations while improving borrower experience and operational efficiency. Processing teams are wrestling with high turnover and backlogs; borrowers are still waiting weeks for decisions and digital first competitors are closing loans far faster and more cheaply on a cost per funded loan basis. Turnover in processing roles is exceeding 35 percent annually, while AI-native lenders are beginning to pull away.¹.

The instinct for many has been to do more of the same, layering new technology onto existing operations, running pilots in isolation or automating individual steps while leaving the underlying operating model untouched. The result, however, is an industry running harder to stand still.

Costs remain elevated, productivity gains stall at the pilot stage and the structural economics of lending continue to deteriorate, leaving most borrowers frustrated with their experience. For many lenders, this has triggered a more uncomfortable question: Can the mortgage business still make economic sense under the current operating model?

Put differently, the mortgage business has become a kind of rabbit hole: Every time lenders sprint to keep up with regulation, demand swings and new technology, margins seem to shrink a little further.

Fortunately, the tools to do something fundamentally different are now here. The convergence of Agentic AI, advanced analytics and outcome-based managed services is giving lenders the means to re-imagine their operating models, not only by accelerating processes but transforming how they connect. This shift is now driving the emergence of AI-first mortgage operating models that combine mortgage automation, intelligent orchestration and AI-powered mortgage operations at enterprise scale.

This article explores the sources of cost pressure, why current approaches are plateauing and the developments enabling the creation of next-generation operating models fit for the future.

Why Mortgage Economics Are Under Structural Pressure

The Rising Cost of Mortgage Origination

The average independent mortgage bank spent USD 12,579 to originate a single loan in the first quarter of 2025³. By the second quarter, that figure had eased to USD 10,965⁴. Taken together with recent cost to originate studies, these figures keep the industry firmly in the USD 11,000–12,000 per loan range, underscoring that simply adding more volume is no longer enough to fix the economics of the mortgage business.

Cost-to-Originate Challenges

Spiraling origination costs are being driven by three forces that are now structurally embedded:

← Swipe →

Mortgage Margin Compression

At the same time, rate volatility is punishing the fixed cost operating models that most lenders still run. In recent years, the 30 year fixed mortgage rate⁵ has often moved by dozens of basis points within a single month, as markets react to shifting macroeconomic data, geopolitical events and policy signals. This level and speed of movement mean pipeline assumptions break and fall through rates spike, with lenders absorbing the full cost of idle capacity because their operations cannot flex with the wider landscape.

Crucially, none of these forces is cyclical. They represent a structural reset in how mortgage economics operate: More volatility, higher fixed costs and rising expectations from regulators, investors and borrowers. Waiting for the cycle to turn is no longer a viable strategy.

A divide is also emerging between lenders who are harnessing emerging technologies versus those who aren’t. Industry surveys indicate that a growing share of mortgage lenders are piloting or expanding the use of AI and machine learning in their operations.⁶

Why Traditional Mortgage Transformation is Stalling

From AI Pilots to AI-first Mortgage Operations

A natural response to these pressures has been to bolt new technology onto existing operations, automating individual steps, running pilots in isolated pockets of the business or adding digital tools to workflows designed for a different era. While impactful, this efficiency focus will only take firms so far.

That’s because the true constraint is not the technology, but the operating model beneath it. Core data remains split across origination, point of sale and servicing systems that cannot interoperate. Pilots show promise in controlled environments, then fail when exposed to greater complexity. And leadership teams demand enterprise wide ROI from initiatives that lack the foundations to deliver it.

Local efficiency gains can buy a few quarters of relief, but they do not change the structural math of the business. They still sit on top of fixed-cost capacity, fragmented data and manual exception handling, which means cost per loan, cycle times and breakage rates remain stubbornly high when volumes or rates move.

The lenders pulling ahead are approaching the problem differently. Instead of layering new tools onto legacy workflows, they are re-building fundamentally different operating models where technology serves as the core infrastructure around which everything else is organized, and decisioning, workflow and execution are designed together.

Doing so means decision intelligence is embedded across pricing, lead scoring, fraud detection and income analysis, while workflow orchestration manages multi-step processes autonomously, across the entire enterprise. Crucially, new approaches also enable human effort to be concentrated where it adds the most value: Judgment, relationships and oversight, rather than data processing and manual reviews.

The Role of Agentic AI in Mortgage Lending

Moving Beyond Point Solutions

The emergence of Agentic AI is making this shift viable. Agentic systems can plan, execute and adapt complex workflows without continual human instruction. In mortgage operations, AI now goes beyond single-task applications like document extraction to orchestrate the full journey, gathering inputs, cross checking sources, identifying inconsistencies and producing outputs on its own.

In practical terms, this looks less like a futuristic assistant and more like a team of specialized digital processors that can pick up a loan file, understand what needs to be done next and move it forward without waiting for a human to push the next button.

Intelligent Workflow Orchestration

The impact is tangible across the mortgage lifecycle:

← Swipe →

These once hypothetical achievements have been rapidly embraced as differentiators by industry players. Rocket Mortgage’s Rocket Logic platform now closes loans 2.5 times faster than the industry average and recently launched fully digital pre-approvals with no loan officer intervention⁸. Elsewhere, Better.com’s Tinman engine underwrites in a median of 2ss minutes and 24 seconds against the industry’s 21-day norm, cutting cost-to-originate by more than 40 percent.^9,10

Re-building Mortgage Operating Models for Scale

What lenders face is not a technology procurement decision but an operating model decision. Those leading the way are identifying the right tools and using them to re-wire the way loans are originated, processed, underwritten, closed and serviced while continuing to serve borrowers every day under existing standards.

Rather than simply running harder, it marks an entirely new course. Whether built in-house or through partnerships, bringing this shift to life requires three capabilities delivered together:

← Swipe →

Variable-Cost Mortgage Operations

For executives, this is where the AI-first mortgage operating model becomes real. Technology is not only deployed; it is run, maintained and continuously optimized by a specialist partner whose incentives are aligned with business outcomes.

The critical point is that these capabilities must function as one system. Consulting is powered by diagnostics derived from operational data. Technology is intelligent automation deployed across the lifecycle. Managed services are technology augmented operations tied directly to business outcomes. This is not a hand off from strategy to implementation to outsourcing, but a single integrated operating model.

Outcome-based Managed Services in Mortgage Operations

How Lenders Can Modernize Mortgage Operations

Realizing this future represents a significant challenge, and most lenders cannot build such operating models on their own. Even for market leaders, re-building a digital native mortgage operating model internally typically requires years of integration effort and hundreds of millions in sunk cost, with no assurance the economics hold before market conditions shift again.

Replicating digital native infrastructure demands sustained investment across data integration, systems engineering and workflow re-design, often while legacy platforms must continue operating under regulatory scrutiny. The data integration challenge alone requires specialist capability to connect systems built decades apart. The talent crisis compounds the problem, making it difficult to build, staff and sustain these capabilities at scale.

Many organizations are making the strategic choice to partner in response, enabling access to the tools, talent and expertise required to overcome these barriers and enter this new era. Effective partners deliver the rebuilt operating model as a unified capability rather than a collection of discrete services, ensuring transformation happens at an accelerated level.

The Future of AI-driven Mortgage Lending

Looking ahead, acceleration will prove integral as the window for action narrows. Lenders who have already re-wired their operations are compounding their cost and speed advantages every quarter. Organizations are actively embedding automation into the infrastructure of the secondary market, raising the baseline expectation for every originator and servicer.^11,12 At the same time, rate volatility is not abating, and every month a lender runs a fixed-cost model against variable demand, the margin erosion deepens.

Promisingly, the economics of outcome-based managed services — variable cost, measurable from day one, instantly operational and aligned with external factors — mean this is no longer a leap of faith. Instead, it marks a necessary decision with transformative payback.

The future of mortgage lending will not be defined by isolated automation pilots or incremental digitization initiatives. It will be shaped by lenders that embrace AI-first mortgage operating models, intelligent workflow orchestration and AI-powered mortgage automation as the foundation of scalable, future-ready lending operations.

References

1. Mortgage Bankers Performance Report – Quarterly and Annual | MBA

2. 2025 Updates to the Cost to Originate Study | Freddie Mac

3. IMBs Report Slight Product Losses in First Quarter of 2025 | MBA

4. IMBs Report Production Profits in Second Quarter of 2025 | MBA

5. Homeowner Mortgage Rate Breakdown | CNBC

6. AI in Mortgages | American Bankers Association

7. Machine Learning Can Save Mortgage Originators Up to $1,500 per Loan | Freddie Mac

8. Rocket Companies Introduces Rocket Logic AI Platform to Make Homeownership Faster and Easier | Rocket Companies

9. Better Announces First Conversational Credit Decision Engine in ChatGPT with OpenAI | Business Wire

10. Better Mortgage’s AI Revolution: How Better is Empowering Loan Officers, Not Them | HW Media

11. Freddie Mac Issues Updates Mandating AI Governance | National Mortgage News

12. Fannie Mae Establishes New AI Governance Guidelines | National Mortgage News