By Lakshya Jain, Director, Mortgage Technology, Annaly Capital Management
I remember when financial services were a place you had to go to. You would walk into a bank, fill out a bunch of forms, and then wait for someone to tell you if you qualified for a loan. That is not how it works anymore. Now you can get a loan from your Shopify dashboard or from the checkout button at your favorite store. You do not even have to go to a bank.
I have been working on underwriting and risk systems for mortgage lending for several years. Specifically, I have been working on Non-QM lending, which is a type of lending that uses alternative ways to verify income, including bank statement analysis and exception-based workflows that sit well outside standard agency guidelines.
What I have learned from this experience is that the big change with embedded finance is not just that financial services are available everywhere. It is about who makes the decisions, who takes the risk, and who is responsible when something goes wrong. Inside a regulated institution, those questions have clear answers. Distributed across platforms, APIs, and sponsor banks not designed to hold joint accountability, they become complicated fast.
Why Embedded Finance Is Growing Rapidly in 2025
The market for embedded finance reached over $148 billion in 2025 and is growing at a rate of over 30 percent per year. This is not just because fintech companies are trying new things. The economics are genuinely compelling.
When you embed a product into a platform, you can capture the customer at exactly the right moment. Research from McKinsey suggests this can increase customer lifetime value by two to five times, while also reducing abandonment and capturing interchange on every transaction.
In Non-QM mortgage lending, we saw this same thing happen years ago. The companies that grew the fastest were the ones that brought the credit decision into the loan officer’s workflow. This made it easier for the customer and reduced the number of people who gave up on the process. But it also created downstream complications when compliance accountability got distributed across systems that were not designed to share it.
Banking-as-a-Service (BaaS) in Embedded Finance: What Is Working and What Is Not
The companies that are winning in embedded finance are not the ones with the most elegant customer experience. They are the ones that have invested in the backend systems, compliance middleware, and audit infrastructure. This is what makes their decisions defensible.
The companies that are doing well are the ones that understand the customer workflow well enough to embed finance at a moment of genuine need. ServiceTitan embedding construction financing through Wisetack and Toast offering working capital to restaurant operators are good examples of this. In Non-QM mortgage technology, we learned this years ago. When you integrate finance deeply into the workflow, compliance stays traceable over time.
What Non-QM Mortgage Systems Teach Us About Embedded Finance
Non-QM lending exists because the standard rules for underwriting do not work for a lot of borrowers. These are people like self-employed business owners with strong cash flow but irregular W-2 income, or real estate investors whose debt service coverage ratios tell a better story than their tax returns. They need a decisioning framework that can evaluate alternative income evidence and document every step in a way a regulator can follow.
When I was working on Non-QM lending systems, I learned that the data ingestion and analysis pipeline has to be accurate and auditable. This is because the loan decision depends on twelve months of bank statements, not a pay stub. The workflows have to be exception-based, which means an underwriter can override a model output but has to justify that override in the decisioning record. Non-QM origination often also requires a credit decision while the borrower is still in the application session, meaning the whole evaluation chain has to move fast without sacrificing the audit trail.
What most embedded finance discussions miss is the capital markets layer. In Non-QM lending, every loan feeds into a securitization pool. Investors in non-agency RMBS make capital allocation decisions based on the integrity of the underwriting process behind each loan in that pool. A system that cannot produce a clean, auditable decisioning record does not just create a regulatory problem. It creates an investor confidence problem that travels up the entire capital stack. Non-QM mortgage technology already operates at that level of accountability. Embedded finance platforms will face the same reckoning as their credit products attract institutional capital.
Embedded Finance Risks and Compliance Challenges Every Platform Must Prepare For
The regulators are paying attention to bank-fintech partnerships. In 2024, Blue Ridge Bank faced enforcement action from the OCC over AML failures in how it administered its BaaS relationships. The consequence is now established precedent: sponsor banks are accountable for compliance failures that happen inside their fintech partners.
The problem is what I call accountability diffusion. When credit decisions happen inside a third-party platform, it is not clear who owns the fair lending obligation. Who runs the disparate impact analysis? Who issues the adverse action notice? Who maintains the model documentation a regulator will ask for?
These are the exact questions regulators have been pressing AI-assisted Non-QM underwriting platforms on for years. The CFPB has been explicit: there are no exceptions to consumer financial protection law for new technology. Using an algorithm makes the explainability burden more demanding, not less.
Even as federal enforcement activity recedes, state regulators are actively filling the gap. Any company embedding lending products should plan for a more active state-level regulatory environment over the next few years.
Why Consumer Trust Is Embedded Finance’s Most Underrated Challenge
The customer experience argument for embedded finance is a good one. It removes friction and meets people where they already transact. But what about trust? What receives far less attention is the question of confidence on the other side of that convenience.
When a consumer accepts a financing offer at checkout, they often have no idea which institution is extending the credit. They do not know what data was used or how it will be handled. The financial infrastructure has become more accessible but less visible.
Data fragmentation is a problem. Customer information travels across origination platforms, middleware layers, and sponsor bank systems before anyone has acted on it. Each handoff creates a compliance risk and a trust risk. The deeper concern is that the system making the credit decision may itself become invisible, spread across a scoring model, a BaaS provider, and a platform with no single entity clearly owning the outcome. A bank that makes a wrong credit call owns it. A distributed credit decision does not have the same clarity.
In Non-QM mortgage origination, borrowers are often surprised by how many systems their application moved through before a decision was made. That opacity does not build trust. Platforms that pair a seamless experience with transparency about how decisions are made will earn a meaningful trust advantage.
The Future of Embedded Finance in Regulated Lending
The lesson from Non-QM mortgage technology is clear. Retrofitting compliance architecture onto a system that was not built for it is expensive and often incomplete. The organizations that struggled most with examination treated governance as an afterthought. Those that sustained growth made those choices at the architecture level from the start.
Companies entering embedded finance in 2025 and 2026 have an advantage. They can see where accountability gaps emerge when compliance is treated as an afterthought. Those infrastructure decisions will determine whether regulatory accountability becomes a moat or a recurring liability.
The next phase of embedded finance will not be defined by how quickly companies can launch financial products, but by how effectively they can build lending systems that regulators can trust at scale. The Non-QM mortgage world built that knowledge through hard experience. The broader embedded finance ecosystem does not have to make the same journey to understand it.
About the Author
Lakshya Jain is Director of Mortgage Technology at Annaly Capital Management, where he leads AI-driven underwriting systems and origination infrastructure for Non-QM mortgage lending. His work focuses on alternative income analysis, credit decisioning platforms, and compliance architecture in regulated lending environments. He can be reached on LinkedIn at linkedin.com/in/jainlakshya.
