TECH ME HOME TONIGHT

Why Lenders Are Building Their Own Tech

The logic behind building in-house is largely about breaking free from the limitations of legacy providers. Most lenders rely on vendor platforms like Encompass or Calyx, but, while widely used, these systems can be expensive, difficult to customize, and slow to adapt to changing compliance or business needs.

And the numbers reflect that. Over 90% of lenders rely on at least one commercial LOS platform for loan applications. With ICE Mortgage Technology’s Encompass holding about 50% of market share and Black Knight’s Empower around 10–15%, together they power roughly two-thirds of all U.S. originations, according to the STRATMOR Group. In spite of the high usage, though, less than 25% of lenders report satisfaction. 

Encompass was originally developed by Ellie Mae, a company founded in 1997 to bring automation to the mortgage origination process. Since its acquisition by ICE in 2020, the platform has evolved into a cloud-based, end-to-end digital mortgage system — but not without problems.

According to a 2025 MarketWise study, loan boarding (the process of transferring a funded mortgage into a lender’s or servicer’s system for ongoing management) remains one of the most time-consuming and error-prone steps in the mortgage lifecycle, with errors affecting up to 2.77% of loans and missing data impacting more than 3.5% post-boarding. Even with modern tools, originators and servicers often operate in separate systems, leading to duplicated work, redundant audits, and an average of 70 minutes spent per loan on intake. 

Loan boarding exposes the friction between systems. For example, Encompass may handle the origination side well, but if it can't hand off that loan cleanly into servicing, lenders still face rework, delays, and regulatory risk — the very problems tech is supposed to solve.

Of course, that fragmentation isn’t limited to servicing. “Very few companies are able to leverage the newest technology,” says Leah Price, former head of financial technology at FHFA and now a senior executive at Better. “The technology is fragmented. It’s very old. And it hasn’t allowed for full integration of AI.”

In her previous role, Price studied technological advancements across the mortgage industry, where she saw firsthand how legacy systems and siloed infrastructure prevented the kind of end-to-end automation that modern tech tools require. 

In moving to Better, the choice was deliberate: to join a company that not only built its own technology stack but also had the scale, capital, and leadership commitment to actually use it — without being hamstrung by legacy infrastructure or vendor dependencies.

Better’s solution was to build a unified platform (Tinman AI) designed to serve both the borrower and the internal operations team. Atop that platform sits Betsy, a generative AI-powered voice bot and chat assistant that can answer borrower questions, draft emails, and support loan officers with file status updates and compliance-appropriate prompts.

“She can banter,” Price said, “but she’s very customized to the rules of originating a mortgage.”

For companies like Better, building technology internally offers operational efficiencies and new business opportunities. “We've iterated with this system over the last 10-plus years that has enabled our loan ops staff to be extremely efficient,” Price said. “And now we're really well positioned to offer that to other lenders — something that wasn't previously available to many.”

And Better isn’t alone. Sun West Mortgage likewise began its tech rebellion early. In 2018, the company launched AngelAi, a homegrown platform designed to act as personal assistant, loan officer, and even credit advisor — all in one. Built by sister company Celligence International, AngelAi automates loan processing, assists with compliance, and helps borrowers improve credit scores through automated dispute resolution.

More than that, though, it was built to feel human. AngelAi uses “empathetic technology” — tracking tone, typing patterns, and behavior to tailor responses and promote fairness in lending. It offers financial guidance, supports real estate agents with SEO-boosted listings, and runs 24/7, never taking a break.

“There’s no carrot and no stick,” CEO Pavan Agarwal told NMP in a 2024 interview. “But because AngelAi learned, she was promoted.”

Better and Sun West built proprietary platforms tailored to their internal workflows — not initially for resale, but to gain a strategic edge. At Better, that meant streamlining the entire borrower journey through a fully integrated, AI-native system. At Sun West, it meant deploying Morgan (the early iteration of AngelAi), an in-house assistant designed to supercharge communication.

For Matthew VanFossen, the goal wasn’t branding or buzz. It was fixing what wasn’t working. Big POS, the white-label frontend developed by Mortgage Automation Technologies, similarly began as an internal project at Absolute Home Mortgage during the early months of 2020. Built atop Encompass, it was designed to give lenders more control over how loan data is collected, displayed, and routed without replacing the LOS itself. 

“When I look at point of sale, I think that our customer lives in there,” VanFossen said. “Our loan officer works inside of there. And our real estate referral partners work inside of there.”

What started as a custom fix for internal inefficiencies evolved into a commercial product now used by more than 35 independent mortgage companies. Its design emphasizes transparency, modularity, and speed — a direct contrast to the rigid, slow-moving vendor tools VanFossen had grown frustrated with.

Why Most In-House Projects Fail

Despite these success stories, the track record for in-house mortgage tech is overwhelmingly poor. Rick Roque, a former Silicon Valley engineer turned mortgage executive, has spent the last two decades advising lenders on tech strategy, acquisitions, and system integrations. He’s seen the pitfalls up close. “Probably the biggest fool's gold has been mortgage companies attempting to build their own loan origination system,” Roque says. “Over the last 15 years, companies have wasted tens of millions of dollars … and they fail.”

Roque cites several reasons for these failures: lack of capital, absence of experienced product managers, and the immense technical debt that accrues when building complex systems in a highly regulated industry.

“They don't have the capital to keep up,” he explains. “You really need to have a product manager mindset — someone who understands user experience, competitive requirements, and regulatory trends. Most mortgage companies don't know how to hire for that.”

In truth, there is often little overlap between loan officers and tech wizards, with lenders struggling to bridge the gap between people who know how mortgages work and people who know how software should be built.

Originators understand borrowers, guidelines, and turn times, but they’re not thinking in sprints or scalable architecture. Meanwhile, engineers might be brilliant at writing code, but they often underestimate just how complex, Byzantine, and regulation-heavy mortgage really is. The result is a communication breakdown where product decisions get lost in translation and the final tool ends up serving no one particularly well.

At Better, however, that gap is narrower by design. Offering a totally online mortgage experience means that product, engineering, compliance, and operations are forced to work in lockstep from day one. “This company attracts people who want to work on hard problems, who want to work with really smart people, and who want to be able to make a difference quickly,” Price said. “That’s not everybody. It’s very uncomfortable for some folks who are used to being in a more traditional environment.”

Leah Price echoed the challenge of maintaining momentum post-launch. “As you build more and more stuff in-house, all of that tech debt accumulates,” she said. “Then it's hard to move quickly and pivot when you decide you want to do that.”

One of the clearest cautionary tales is Interfirst Mortgage, which, after shuttering in 2017, re-entered the market in 2020 with an ambitious plan to relaunch as a tech-driven lender. The company raised $175 million and built its own AI-powered LOS. But by the end of 2021, Interfirst had laid off dozens of loan officers and retrenched. Their platform simply hadn’t delivered results fast enough to justify its cost. 

Another example is Open Mortgage, which once relied on an in-house LOS until new compliance rules under TRID rendered it unworkable. The company attempted to replace it with a vendor solution in early 2015, but the first rollout failed. Under pressure from regulators, they scrapped the project and started over with a second vendor, racing to implement a compliant LOS in just 50 days.

In both cases, the firms were forced to walk away from years of work — and likely millions in sunk costs. It’s a painful reminder that building software in-house offers control, but at a steep price.

The real challenge, as Comidor observes, isn’t the code, necessarily, but everything around it. “You need to decide on teams, supervise them, and, in some scenarios, even micro-manage them,” they write. Between high upfront costs, long ramp-up times, and the constant pull on executive attention, the effort often outweighs the return. The tech doesn’t necessarily fail because the code is broken. It fails because no one has the time, budget, or stamina left to wield it.

What Makes A Build Succeed?

The companies that do manage to build effective in-house tools tend to follow a few hard-earned principles; lessons learned not from whitepapers, but from shipping code in the middle of a pipeline backlog.

Still, these successful builds are the exception, not the norm. Across the industry, fewer than 10% of lenders have deployed AI tools at scale according to data from Fannie Mae, and even fewer have pursued fully custom LOS or POS platforms. 

One reason is cost: while tools like eClosing have been shown to reduce transaction costs by up to $444 per loan, fewer than 14% of lenders have adopted eClosings across 80% or more of their pipeline, largely due to upfront investment and integration complexity.

That gap between ambition and execution explains why internal builds are so rare — and why, when they work, they matter. The companies that succeed aren’t trying to win awards or land on stage at a conference. They’re solving their own problems first.

First, they build for themselves before they build for others. Both Better and Big POS started as internal tools, solving internal pain. There were no press releases, no client roadshows, just ops teams and originators using the product day after day. That tight feedback loop created something a vendor demo can’t: real-world validation.

“We experiment on ourselves,” VanFossen said. “Sometimes it goes really bad.” But when it goes well, it means you’ve got a product that’s actually lived through the workflow it’s supposed to support.

Second, they embed compliance and risk mitigation early. Baked in rather than added on. Mortgage isn’t a space where you get to ask for forgiveness later. “You need to frontload all of the risk management and the mitigants,” said Price. “It’s not, ‘Let’s build this cool thing and then figure out how to make it compliant.’”

That means involving legal and QC from day one. It means structuring your product sprints around regulatory deadlines and testing edge cases like they’re use cases. And it means accepting that some features won’t launch on time — because in mortgage, “done” isn’t done until it’s audit-proof.

Third, they maintain a narrow focus. Ambition kills more builds than bad code ever will. Better, for instance, chose not to build its own document capture or processing tools. Instead, they license vendor tech for commodity functions, and reserve their engineering talent for what matters most: customer experience, pricing logic, and internal workflow automation. “We focus on what we can turn into valuable intellectual property for ourselves,” Price said. “And what we probably won’t.”

Fourth, they get leadership aligned early and stay aligned throughout. In failed builds, it’s common to hear that the CEO “approved the budget,” but wasn’t involved in the roadmap — or that ops was told to use the system but wasn’t asked what they needed from it. Successful builds don’t work that way. They’re driven by senior leaders who treat tech as infrastructure, not accessories. Product decisions aren’t punted down. They’re owned from the top.

And finally, they obsess over adoption. It’s not enough to ship the code. People have to want to use it. That means simple interfaces, clear training, and continuous iteration based on user feedback. If your loan officers are still keeping sticky notes on their monitors after rollout, your build didn’t land.

“Adoption is everything,” said Roque. “You can build the best system in the world, but if it doesn’t make the LO’s life easier by day three, they’re going to ignore it.”

Taken together, these principles represent the bare minimum for surviving an in-house tech build. Companies that forget this don’t just end up with bad software. They end up with broken workflows, frustrated staff, and a very expensive reminder that building tech is only worth it if it works in the wild.

Looking Ahead

Even with the potential for failure, the appetite for in-house tech is evolving. Fast.

For example, AI is no longer a hypothetical. At Better, it’s baked in. Betsy, rather than your standard pre-fab chatbot, is a compliance-aware virtual assistant who speaks to borrowers, responds to loan officers, and integrates with internal systems in real time.

“She can say things like, ‘Here’s what’s missing in your file,’ or, ‘Here’s what you need to do next,’” Price explains. “She can even initiate contact with borrowers and help ops staff draft templated communication.”

What’s more important than Betsy’s personality is her plug-in, Tinman AI. It’s what makes Better’s system fast, adaptive, and, at least for now, difficult to replicate.

Roque believes this kind of automation is going to explode over time, radically reshaping headcount, industry-wide. “You’re going to see 120,000 people laid off in mortgage,” he says. “Not because they’re not good at their jobs, but because AI will be doing them.” Underwriters, processors, marketing staff — anything repetitive is at risk.

Additionally, Roque predicts a murky future for the legacy platforms themselves. “Five years from now, my view is that we’ll be in fullscale automation like AI automation and personalization. Tactically speaking, the loan origination platform won’t need to exist,” he says. ““There won’t [be a] loan application because the loan application will just automatically be assembled by AI through just various questions being asked to the consumer. Loan officers 5 years from now will not be logging into Encompass.”

Price is more optimistic about the future. She envisions a world where artificial intelligence streamlines the tedious, time-consuming parts of the mortgage process — chasing W-2s, checking statuses, drafting boilerplate emails — so that humans can focus on what they do best: connecting.

“In five years, all of us are going to be a lot more comfortable with AI tools,” she said. “It just makes the excruciating parts flow better.” As the machines take over the paperwork, the human edge becomes more clearly defined and more valuable. “The most successful people will be the best at soft skills,” she said. “They will really be able to tap into things like empathy and one-on-one coaching and making consumers feel more comfortable with the human aspects of the transaction.”

But getting to a utopian paradise of soft skills is another matter altogether because building mortgage tech in-house is still a gamble.

It demands capital, vision, compliance fluency, and a willingness to break things in full view of your sales floor. Most lenders won’t do it. Some will try and fail. 

But a few will pull it off, and in doing so, will redraw the map for everyone else.