From Ai Curiosity To Operational Transformation: Michael Vandi On How Mortgage Lending’s Relationship With Ai Is Evolving

AI in mortgage lending has evolved rapidly over the past two years, shifting from curiosity and experimentation to enterprise-wide operational transformation. Michael Vandi, founder and CEO of Addy AI, has spent the past several years working with mortgage lenders as mortgage AI adoption has evolved from experimentation to enterprise AI mortgage implementation.

Vandi saw firsthand how AI in mortgage lending emerged as one of the industry’s earliest and most consistent technology shifts. He shares why lenders embraced the technology so quickly, how AI conversations have shifted from experimentation to implementation and why trust, more than the technology itself, is becoming the defining factor in successful mortgage AI adoption.

HousingWire: Mortgage professionals were among the earliest adopters of AI. Looking back, why was the industry such a natural fit?

Michael Vandi: When we first looked at our users, mortgage wasn’t actually the largest group using the product, but it stood out because of how consistently people used it. Other industries, like e-commerce, had very different use cases from one business to the next. Mortgage was different.

The products are largely the same across lenders. Everyone is working with similar loan products and workflows, which has made it much easier to identify recurring problems that AI can solve. That consistency made it feel like an enterprise opportunity where we could go very deep, rather than trying to build something that worked a little differently for every customer.

That’s what convinced us to focus on mortgage and eventually build a team with deep mortgage expertise alongside the AI expertise.

HW: How have conversations with lenders changed over the past two years?

MV: They’ve changed dramatically. Early on, the conversations were mostly, “AI is cool. Let’s see what it can do.” Lenders were reading about companies using AI and felt like they needed an AI strategy because everyone else was talking about it.

Now the conversations are much more operational. Lenders come in with a clear understanding of the workflows they want to improve. They’ll say they have a certain number of processors, handle a certain loan volume and believe parts of that process can be automated.

The challenge isn’t recognizing the opportunity anymore. It’s implementation. As mortgage AI adoption has matured, lenders are focused less on whether to use AI and more on how to deploy it successfully within existing workflows.

That’s how engagements often expand. A lender might come to us wanting AI to validate closing documents, but once those agents are in place, they realize the same technology can scrub files, validate pre-underwritten loans and automate other steps throughout the workflow.

HW: You’ve described mortgage AI adoption as a trust curve rather than a learning curve. What do you mean by that?

MV: The learning curve really isn’t the issue. Our experience has been that people can learn to use AI tools very quickly. In many cases, they’re easier to use than traditional mortgage software.

What changes over time is trust. At first, users want to verify everything the AI does. They’ll let AI complete the first pass, but they’ll carefully review every recommendation before moving forward. As they continue using it on live files and see it deliver reliable results, they begin to trust it with more responsibility.

That’s why I think AI adoption is really a trust curve. The technology isn’t difficult to learn. Organizations gradually become comfortable relying on it as it consistently proves itself.

HW: As lenders expand AI across their organizations, what separates companies that make real progress from those that stay stuck in pilot projects?

MV: The organizations making progress are approaching AI through operational workflows instead of treating it as a standalone technology. The lenders we’re working with already have a thesis about where automation can create value. They know the problems they’re trying to solve. Once AI demonstrates success in one part of the process, they’re willing to extend it into adjacent workflows.

We’ve seen lenders begin with one operational workflow, such as document validation, then expand AI into adjacent processes like file review and pre-underwriting once confidence grows. Instead of deploying AI for a single task, they began to consider how AI in mortgage lending could improve the entire loan lifecycle.

HW: Looking ahead, how do you see AI reshaping mortgage operations over the next few years?

MV: AI is already changing how we work internally, and I think the same shift is coming to mortgage. Instead of AI helping with individual tasks, organizations are beginning to hand over entire responsibilities to AI agents. That changes how people think about work and where they spend their time.

We’re already seeing operational efficiency create new opportunities. We’ve seen lenders become significantly more efficient in loan operations, enabling them to launch new loan products much faster than before. Work that traditionally took months, or even years, can now move much more quickly because operational bottlenecks have been removed.

Mortgage is also a very people-driven business. Many experienced underwriters have spent decades in the industry, and it’s fascinating to watch them discover what AI can do. The technology isn’t replacing their expertise, but it is changing how those roles operate and how responsibilities are shared between people and AI.