How to rebuild the $13 trillion mortgage industry: become a servicer first for six years, then sell them the software
The industry won't trust untested systems. Valon first operated as a mortgage servicer for six years to prove its technology, then sold it back to the industry as software.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
A $13 trillion industry that never caught up to 2000s-era technology
According to Linda, mortgage lending is probably among the top three least disrupted industries. Of the $13 trillion in consumer mortgage debt, most runs on legacy systems built before the internet existed. This isn't competing for the last 20% of a market—it's an entire industry that hasn't caught up to 2000s-era technology. That's the scale of the opportunity. Deployment gets harder in proportion: you need to train tens of thousands of employees, completely replace core operating systems, and actually deploy AI in a heavily regulated industry.
Sixty-year-old systems don't understand today's world
These systems were designed in the 1960s, but rules around property insurance and mortgage insurance keep changing while the data models stayed frozen. Bugs don't surface the week they're introduced—they compound for five years before appearing. An escrow account calculated with faulty logic won't fail until the homeowner tries to pay; the error only surfaces then. You can't empathy-talk your way around it if the system has the wrong number; only rebuilding the foundation fixes the experience.
Three paths existed; they chose the most painful one
Three options: build software and sell it to incumbents; acquire legacy service providers and inject new technology; or build a complete service provider and software platform from scratch. Option one doesn't work: regulation is so strict nobody trusts an unproven system, and once you bind to one large customer, you get pulled back into their operating model. Option two: IP gets trapped. Acquired platforms can only be replaced module by module, locking you into someone else's original architecture decisions. Only the third path remained—the hardest but most thorough.
Getting licenses means solving a chicken-and-egg problem first
Many licenses require you to first prove profitability to get the license—New York State and Fannie Mae and Freddie Mac follow this logic. So the team had to target non-QM loans where state licenses covered the ground. But large states like California require federal loan approval authority first. The only loophole: a real estate broker license, but you also have to prove five years in collections, customer service, payment processing, and foreclosure work. Each license class locks into the next; the industry estimates three to five years for the full cycle.
Six pandemic months of reading regulations into code
There's no shortcut to codifying regulation: federal statutes like RESPA, TILA, FDCPA, GLBA, TCPA, plus fifty states' rules on mortgages, collections, foreclosure, privacy, and escrow. Read every page, then abstract into a unified framework baked into the architecture. During the pandemic, he read regulations eighteen hours a day for six straight months, annotating and sketching schematics. That brute force got the work done. Then five to six years of testing to ensure the system ran to spec.
Tripling efficiency turned break-even into 70–80% margins
Servicing is a unit economics game: better technology means higher profit retention. Valon passed savings to asset managers as price cuts for market share. By their math, they're three times more efficient than legacy servicers. That turned a break-even business into one with 70–80% operating margins. Those extra margins funded the urgency they used to land early customers. After signing one large customer, the other giants came calling.
The entire business model in one phrase: everything is servicing
Residential mortgage servicing is the start. One direction is commercial real estate: as a commercial mortgage servicer, you get tenant financials—the raw material for SMB lending. Another direction is hospital revenue cycle management, which is servicing by another name: instead of homeowners, you track patients and claims, getting data equivalent to complete electronic health records. They chose residential mortgages deliberately as the hardest entry point: once it works here, every other industry gets easier.
Deploying AI at scale is about changing people, not technology
Six years ago, this was a technical problem. Today: it's a change management problem. Large servicers have tens of thousands of employees, each with different incentives, layered hierarchy to navigate. Over the past six to twelve months, the team built the muscle for driving organizational change—brutally hard at scale, but they see it as the biggest value lever AI unlocks over the next decade. People who make this work need high autonomy, comfort with ambiguity, empathy for customers, and the ability to align conflicting perspectives toward a single goal.
In their own words · checked verbatim
It's $13 trillion of consumer debt that basically runs on a single incumbent that built their legacy system before the internet was invented.
you can have the world's best customer experience team and the most empathetic call center agent in the world. But if fundamentally you were charged the wrong amount of money because the system of record was flawed, then you can't fix that homeowner experience without fixing the infrastructure.
We actually got the approval for New York in a record time of, I want to say, three years. So it was a record time of three years. And we actually never ended up getting approved to license or to originate mortgages in New York.
So we are about three times as efficient. And so you take this breakeven business and you turn it into sort of a 70, 80% operating margin business.
we actually want to change the industry. And the thing is, is that if you keep it as a servicer, you're keeping all of the technology and the alpha for yourself. And at the end of the day, you didn't actually change anything. You didn't fix the core infrastructure problem.
And it turns out revenue cycle management is just servicing for hospitals and their claims and they're handling their patients.
If you had asked me six years ago, I would have told you, you know, this is a technology problem. What I know today is that this is a change management problem.
Figures
| U.S. residential mortgage debt | $13 trillion | 0:00 |
| Valon's efficiency advantage over legacy servicers | ~3x | 19:50 |
| Operating margin after efficiency gains | 70–80% | 19:50 |
| Valon Mortgage's managed UPB | ~$200 billion | 20:52 |
| New software deals in first 6 months | >$200 million | 33:26 |
| Rhythm loans transferred to Valon | 4 million (~10% of market) | 33:26 |
| New York State license timeline | 3 years (ultimately denied) | 15:43 |
| Full license approval timeline | 3–5 years | 13:43 |
| Management team with 5+ years tenure | 75–80% | 36:33 |
Glossary
- UPB
- Unpaid Principal Balance; the core metric for measuring a mortgage servicer's managed asset base
- non-QM loans
- Non-qualified mortgages; loans that don't meet government standards but have lower regulatory barriers
- champion challenger
- A testing method that runs multiple strategies in parallel and compares results to pick the best performer
- revenue cycle management
- A hospital's process for handling insurance claims and patient billing—essentially servicing for healthcare
How to listen
Entrepreneurs and investors interested in how software penetrates highly regulated legacy industries; product leaders exploring how AI agents work in compliance-heavy scenarios.
The intro (0:00–2:01) is a preview recap that repeats later; skip it.