Joe Dahleen in Rick’s Room

In this episode of Rick’s Room, Rick Grant sits down with mortgage industry veteran Joe Dahleen to explore the evolving role of applied AI in mortgage lending and how lenders can reduce costs through agentic, role-based automation.

Joe has spent nearly three decades working across multiple facets of the mortgage and proptech ecosystem. This breadth of exposure gives him a unique vantage point. Rather than thinking strictly within the confines of the traditional Loan Origination System (LOS), Joe looks at how mortgage opportunities originate outside the mortgage ecosystem—and how new technologies can connect those upstream data sources to lending workflows.

Joe sees digital bots replacing back-office outsourcing models that traditionally charged per processed loan regardless of funding. The goal: Reduce loan origination costs to below $7,000 per loan, where labor typically represents 60%+ of total cost.

Lightly Edited Transcript

Rick Grant:
Hi, and welcome back to Rick’s Room. I’m Rick Grant, and this is a podcast I designed to recreate that feeling we get at the end of a busy conference. We’re on our feet all day, working hard, meeting a lot of people, and we still haven’t seen the person we were hoping to see — until the end of the day, when we’re finally in a comfortable seat, relaxed, and here comes that person across the floor for a short conversation.

In this case, that person is Joe Dahleen.

If you’ve been around the industry at all, you know Joe is a mortgage industry visionary. And I say that because he has literally worked on all sides of this business. He’s seen technologies come and go, and I always love catching up with him because he usually tells me what’s coming right around the bend.

Joe, thanks for being with me.

Joe Dahleen:
Yeah, great. Awesome. Thanks a lot, Rick.

Rick:
It’s been a while since we’ve talked, and I’m sorry it’s been so long because I love visiting with you. Catch me up. What have you been working on?

Joe:
Yeah, you know, I did a lot of work in the proptech area for the last couple of years, which I thought was very entertaining and really challenging. I had a couple of great friends who launched companies, and that got me thinking outside the box a little bit more. That’s been a lot of fun.

I’ve also been working on a couple of AI projects for folks I did consulting for over the last few years. That’s been fun too.

Now I’m really focused on the applied AI side. I really think that’s the next thing that’s going to propel our industry toward lower cost and more efficiencies for lenders. That’s where I’ve been studying most of the time over the last three or four years.

Today, I feel more like an AI solutions engineer — even though I’m in charge of strategy and sales. I’m really not selling. I’m helping people understand that this is an opportunity, and where it might fit in their tech stack. Honestly, I feel like I do a lot of free consulting.

So that’s what I’ve been focused on lately — honing down the mortgage process to make it understandable for folks, educating people on the process, and seeing whether it actually fits. And I’ll be the first one to tell them if it won’t.

Thinking Outside the Mortgage Box

Rick:
I want to talk more about what it means to be an AI solutions engineer, because that sounds like something we desperately need. But first, I want to go back to something you said — getting outside the box.

It’s been almost 30 years since I met you at an industry conference. I may have thought you had a box for about 30 seconds, but I’ve never known you to think inside it.

What’s the mindset that’s allowed you to do so many different things in this industry over three decades?

Joe:
I think it’s having opportunities to focus on different parts of the process — and understanding where the data actually comes from, and how people are doing things before we even begin the mortgage transaction.

Recently, you’ve seen big transactions happening where real estate and mortgage are becoming one. Mortgage insurance and title are really the only ways lenders can make money nowadays. That’s been a significant shift.

I’m part of a large proptech group from my early days at Zillow, from 2007 to 2012. That’s when I got into consulting, serving on boards, and helping entrepreneurs build infrastructure. Having those blinders taken off gives you vision into what the future could look like.

I’ve served on boards with FormFree, Prudent AI, and Plunk. Knowing what these solutions are doing and how they can be leveraged inside mortgage puts me in a good position to connect the dots.

When I say “get out of the box,” I mean get out of your way in the mortgage area. Today, we’re still working with LOS systems that have been around for 30 years. We’ve packed all these solutions around them, but people don’t always see the opportunities while they’re staring at their screens doing “stare and compare” all day.

Having a fresh mentality and understanding how opportunities originate outside the mortgage space — before they become a mortgage — is one of my strong suits.

Silverwork and Role-Based BO

Rick:
Is that what Silverwork does? You’re in charge of strategy there now. Tell me about the company.

Joe:
It’s a unique approach. It’s not a startup — it’s been around since 2017. It’s led by founder and CEO Mike Kasein, who’s been in the business a long time. He helped take UWM from $1 billion a year to $1 billion a month, then worked at Homepoint before starting Silverwork.

When I saw what he was doing with role-based, agentic bots, it made a lot of sense. I didn’t see many people attacking the problem that way.

Basically, I focused on how these bots could automate entire roles. As Garth Graham has said, the average mortgage costs about $11,000 to originate, and around 60–65% of that is labor. Another chunk is cure costs on the back end.

If I can help lenders get under $7,000 per loan, I feel like I’m doing my job. That makes them more competitive.

The bots operate 24/7 for originators. They change how originators work with clients.

Avoiding the Trough of Disillusionment

Rick:
A lot of lenders want automation. But when they try AI, they don’t see the results they expect. It’s harder than they think. How do you guide them through that?

Joe:
Our strategy is different. This is not an IT lift — it’s a subject matter expert lift.

We develop and customize bots based on their job aids and workflow. I analyze their job aids, understand their tech stack, and determine whether a bot fits. I know the LOS platforms. I’ve used them myself — I’m still a licensed originator.

When lenders realize it’s not an IT project, it becomes much simpler.

A Different Economic Model

Rick:
That has to be attractive, especially given the horror stories around failed tech implementations. What feedback are you getting?

Joe:
It’s unusual because you’re not working through CIOs and CTOs trying to push something in.

Some tech leaders say, “We’ll just build it ourselves.” I say, “Great — go for it.” There are talented people out there. But this isn’t easy.

Another difference: there’s no setup or implementation fee. The founder believes in putting the bot out there and billing only when the loan funds. If it doesn’t fund, you don’t pay.

That’s refreshing.

In the past, lenders used BPO services onshore or offshore. You paid every time a loan was processed — whether it funded or not.

Now we’re doing digital process outsourcing.

“DPO is the new BO.”

We give the bot a seat inside Encompass, like you would a back-office employee. The bot has a name. You can email it.

You can say, “Hey, that borrower isn’t buying the $450,000 house — it’s $425,000. Rerun the file and send an updated pre-approval letter.”

The bot does it — even on a Sunday night.

Looking Ahead

Rick:
Where does this go from here?

Joe:
We’re seeing heavy investment in this space. But financial services is heavily regulated. You can’t just plug in ChatGPT and start building agents that handle sensitive data.

We’ll see private LLMs — closed environments trained specifically for mortgage.

GenAI is great for marketing and education. But once the file lands in your POS and LOS, that’s applied AI.

We’ve used applied AI for decades — DU and LP are examples.

The future is agentic AI performing repetitive tasks — all the “stare and compare” work. But lenders do things differently, so bots need to be trained to each workflow.

It takes time. These private models have to catch up to your process.

We don’t use GenAI with consumers. We focus on the back end — analyzing findings, credit data, reports — and supporting licensed loan originators.

Rick:
I love it, Joe. In a heavily regulated industry, lenders can’t just build something and hope it’s compliant. Someone with your experience building these systems makes sense.

We’ll keep this short, but I want you to promise you’ll come back as lenders figure out this shift from outsourcing with people to outsourcing with technology and AI.

Joe:
Absolutely.

Rick:
We’ve been talking with Joe Dalene, Managing Director of Strategy and Sales at Silverwork Solutions. Joe, thanks for being here.

Joe:
Always great to see you, Rick. Thank you very much.