How Valon Rebuilt a $13 Trillion Industry From Scratch

1. Key Themes

Become the Operator First, Then Sell the Software

Valon rejected both "sell software to incumbents" and "roll up incumbents," instead becoming a licensed mortgage servicer itself and building the system of record from nothing. Being the operator gave them proof, margin, and a forcing function for product quality. Linda Du explains why each alternative failed: selling software "doesn't work for mortgage because it's so heavily regulated that no servicer is going to use a de novo system," and a marquee customer would mean "you wake up 10 years from now and you rebuild the legacy system with like a really pretty UI" [00:11:24]. The servicer path "is the most painful and it takes the longest" [00:10:54], but it let the software "evolve as the servicer was growing as well. So it's like we started with one loan... and ended up with close to a million" [00:11:53].

Using Your Own Software at Scale Is the Ultimate Credibility Moat

In a regulated industry, trust comes from having taken the operating risk yourself. Valon's servicer passed state exams and agency audits on its own platform. Linda Du: "you get safety because we built a servicer that actually used the technology and proved for the last six seven years that it's gone through every state exam every Fannie Mae audit Freddie Ginny you know all of the different sort of compliance hoops" [00:19:16]. She adds: "it's safe because we actually took the operating risk of running it running our own servicer on this system" [00:19:45].

Efficiency as the Go-to-Market: 3x Efficiency Turns a Break-Even Business into 70-80% Margins

The servicer's economics were the product demo and the wedge. Linda Du: "we are about three times as efficient and so you take this break-even business and you turn it into sort of a 70-80% operating margin business and that margin is what we used initially to create that urgency" [00:20:39]. She also notes the buyers' psychology: "all of these asset managers who own these mortgage assets they are economic animals and so at the end of the day it's like when you lower price it's like you can get the market to capitulate" [00:18:47].

Regulation as a Compounding Moat (Licensing Takes 3-5+ Years)

The licensing and regulatory encoding work that looked like a burden became the barrier to entry. Andrew Wang notes the process "at the time was very painful and now it's a very nice mode" [00:12:53]. Andrew (Valon's CEO) says "Most people forecast this type of endeavor taking three to five years at the minimum... because they all stack on top of each other. You can't do one without the other" [00:15:02]. Some licenses even require profitability first: "they tell you you actually need to be profitable before you get license" [00:13:05].

Legacy Data Models Are the Real Source of Bad Customer Experience

Customer pain is downstream of a flawed system of record. Linda Du: "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" [00:09:50]. Andrew (CEO) gives the example of a deceased borrower's heirs being denied history because systems "only have what is going on at any given point in time" [00:06:25].

AI Moves Automation from Deterministic Tasks to the Long Tail, Enabling "Mini Armies of Agents"

Clean structured data is what makes AI useful in regulated workflows. Andrew (CEO): "before you could only automate things that were clearly deterministic... what you can do today is completely different not only can you automate a much much higher percentage of tasks the long tail the really complex scenarios but... you can do champion challenger type tactics" [00:25:59]. Operators will be "operating little mini armies of agents" [00:23:28]. Servicers move "from one that is really commoditized to one that is much more specialized and really flavor driven" [00:25:28].

Continuous Learning Infrastructure Is the Hard AI Problem in Complex Domains

Beyond a harness and vector databases, the differentiator is customer-specific continuous evals. Andrew (CEO): "building a system that allows you to continuously run the evals pull the data out make sure you're actually continuously improving the model in the direction that is aligned not just with what servicing or the way that servicing should be done generally but customer specific" [00:00:13]. In mortgages, correctness matters more: "if you don't know the answer it generally says I don't know versus let me give you you know some hallucinated answer" [00:28:50].

Servicing Is the Universal Infrastructure Layer

Linda Du frames servicing as the pattern behind many industries: "it's the underlying critical infrastructure that supports basically anything that has money movements some sort of regulation and then an operational component" [00:00:43]. Andrew (CEO) maps adjacent markets: commercial mortgage servicers see tenant financials ("you could kind of do the same thing" as Toast [00:31:08]), and "revenue cycle management is just servicing for hospitals and their claims" [00:00:39].

Enterprise AI Deployment Is a Change Management Problem, Not a Technology Problem

Linda Du: "if you had asked me six years ago I would have told you this is a technology problem what I know today is that this is a change management problem" [00:38:01]. She calls change management "the number one biggest value driver for the next decade" [00:38:30]. The ideal deployer: "high agency high ambiguity high empathy for the customer" who can bridge competing objective functions [00:38:46].

Mission-Driven Talent Beats Compensation, Even Against Frontier Labs

Andrew (CEO): people with offers from "open AI, Anthropic" join because "there's a lot of smart people working on these other problems those will get solved... are you going to be the person who actually solves this problem" [00:36:06]. Retention reflects it: "75 8% of our like management team is filled with people who've been here for five plus years" [00:37:02].

2. Contrarian Perspectives

Don't Sell to Incumbents, Compete with Your Future Customers

Conventional wisdom (Harvey-style) says sell software into incumbents. Valon deliberately competed against its future customers. Linda Du: "we took our servicer and we competed in the market and so that's one way of creating urgency because now you're actually directly competing with your software... with your future software customers and hoping that they'll forgive you down the line" [00:19:45]. The result was a flywheel: "once you get your first big customer to sign up then the other five big guys call you and say hey were you actually serious that you wanted to sell software" [00:20:39].

Walking Away from a Scaled, Profitable Business to Keep the Mission Pure

Valon had a viable $10B+ company in front of it as a servicer. Linda Du: "it is a fully viable path to just build a big servicer there are 10 billion dollar plus companies that do this but... 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" [00:22:04]. They then sold Valon Mortgage to Carrington to focus on software.

Mortgage Math Is "A Finger in the Air," Operational Systems Are the Real Alpha

From the hedge fund world, the sophisticated modeling is less valuable than plumbing. Andrew (CEO), on prepayment modeling: "it's all a finger in the air. Just guess your way through math... You're like, yeah, I think it's going to prepay at X speed" [00:02:38]. The real edge was discovering "it's a big systems engineering problem" where "the money tied... Turns out it never tied" [00:03:06].

The Hardest Industry Is the Best Starting Point

Rather than the easy beachhead, Valon chose the most complex. Linda Du: "we just picked mortgages to be first because it's the stickiest it's actually the hardest and most complicated as well and so you know we thought why not do that one first and then we'll go conquer the rest" [00:33:06]. Andrew (CEO) argues that once you win servicing "you're effectively [in a position where] you have tentacles across the entirety of the business" [00:31:37].

The Next Decade's Scarce Skill Isn't Models, It's Applying AI Inside Real Organizations

Linda Du: "the skill set that people will be hiring for in the next decade is the ability to apply AI. And our team's learning that by doing it in the hardest industry possible" [00:04:23]. This contradicts the view that model access or research talent is the bottleneck.

3. Companies Identified

Valon

Mortgage servicing technology company that built a mortgage servicer and its own system of record ("Valon OS"), now selling the platform to the industry. Mentioned as the subject; it signed "over 200 million dollars of deals" within six months of going to market as software [00:34:24], is ~3x as efficient as incumbents, and secured a transfer of four million loans from Rithm ("New Res" in the transcript), almost 10% of the market.

Rithm Capital (transcribed "New Res"/"Rhythm")

Mortgage and asset manager that confirmed a transfer of four million loans to Valon. Angela Strange: "Rhythm confirmed that they're going to transfer four million loans over to Valen which I think is the largest servicing transfer or it's almost 10 percent of the market" [00:33:26]. Linda Du explains its conviction came from seeing performance firsthand because "Valen Mortgage subserviced New Res Loans" [00:34:24].

Carrington

Acquirer of Valon Mortgage. Angela Strange: "You guys just recently announced the sale of Valon Mortgage over to Carrington and now you're taking that OS and selling it into the industry" [00:21:44].

Harvey

Legal AI company cited as an example of the "build software and sell into incumbents" model. Andrew Wang: "that's been successful in many industries, right? Like Harvey and Law is doing that right now" [00:10:14].

Toast

Restaurant software company used as an analogy for platform-led small business financing. Andrew (CEO): "how like toast and a bunch of people used their platforms to figure out how to do small business financing it's like you could kind of do the same thing if you decide to go through commercial" [00:31:08].

Fannie Mae, Freddie Mac, Ginnie Mae

Government agencies whose audits and approvals Valon passed. Linda Du: "every state exam every Fannie Mae audit Freddie Ginny" [00:19:16].

Soros Fund Management

Where Andrew (CEO) first encountered mortgage servicing problems while investing. Andrew Wang: "your time at Soros and how you decided that mortgage servicing was going to be a really interesting problem to work on" [00:02:01].

OpenAI and Anthropic

Cited as the alternative employers Valon recruits against. Andrew (CEO): "they had offers to go to work at... open AI, Anthropic" [00:36:06].

4. People Identified

Andrew (Valon Co-founder and CEO)

Former Soros investor with a math background, Valon's CEO who also moonlights as one of its most productive engineers. Mentioned for reading every federal and 50-state regulation. "COVID happened I spent 18 hours a day for six months straight just every day sitting there reading regulation and like annotating and basically coming up with the schematics" [00:17:28]. He also went to a licensing officer's mailroom to find a lost New York application: "I went to the mailroom and I identified where the package was" [00:16:09].

Linda (Valon Co-founder)

Valon's co-founder, who describes the market opportunity, go-to-market, and enterprise deployment learning. Mentioned for the strategic framing: "mortgage is probably top three in terms of undisrupted industries that exist" [00:03:55], and the safety-and-urgency framework for software sales [00:18:47].

Angela Strange

a16z general partner who hosts and invested in Valon. Mentioned as the interviewer who framed the market: "Mortgage servicing is a $13 trillion market that still runs in large part on technology designed before the internet" [00:01:01].

Jensen Huang

Nvidia CEO, referenced for a quote. Andrew Wang: "it's really Jensen's quote of it's not like AI that's your competition it's your competition using AI faster" [00:20:22].

5. Operating Insights

Publish the Full Master Plan in the Pitch Deck, Then Execute It Step by Step

Valon's Series A deck was an unusually detailed "60 page PDF of the master takeover plan and then we followed it step by step" (Andrew Wang) [00:21:44]. Linda Du notes "we actually share our series A deck with our" new employees [00:21:44], who respond "you guys are selling this thing this is always what we've been meaning to do" [00:21:33]. Sharing the long-horizon plan internally aligns hires and gives conviction through years of grind.

Structure Pricing as Outcomes: Earn Only When the Customer Earns or Saves

Linda Du: "we structure everything as outcomes oriented and so it's for every dollar that our customer earns or for every dollar that our customer saves that's when we earn as well" [00:34:53]. She adds that "the second that you lose that alignment that's when you're kind of going downhill as a company" [00:34:53]. Combined with an operating "demo room" (the subserviced loans), this removed buyer risk.

Sequence Regulatory Approvals Deliberately Around What You Can Get First

Because major government approvals were out of reach, Valon chose a niche (non-QM loans requiring only state licenses) and sequenced states intentionally. Andrew (CEO): "you have to be really methodical in plotting out what licenses you get and what order" [00:13:34]. Build a staircase of approvals instead of waiting for the final one.

Build Change Management as an Explicit Muscle

Linda Du: "what we've spent the last six to 12 months building is really that change management muscle of just how do you navigate organizations how do you make change happen" [00:38:01]. Hire for both problem-solving and people skills: "you need both" [00:39:15]. Treat each enterprise deployment as a skill to be systematized.

Refactor the Legal Code Into an Abstract Framework Before Writing Product

Andrew (CEO) treated regulation as a codebase: "refactoring the entire legal code base... many legal statutes are derived off of different states and so you sit there basically figuring out how to create an abstract framework that handles all of it and then you put it into your architecture" [00:17:28]. Configurability then lets servicers adapt on the fly when "regulations change, it might be some big fire happened" [00:09:01].

6. Overlooked Insights

The Servicer Relationship Is an Unused Monthly Customer Touchpoint, and That's the Real AI Prize

Mentioned in passing, Linda Du describes turning servicing into a recurring relationship with every homeowner: "I'm going to call them every single month I'm going to build a relationship with them I'm going to make sure I have a pulse on what's going on with them every single month and that's never been done before and that's not how people think of servicing" [00:32:36]. Combined with commercial servicers seeing tenant financials and RCM touching electronic medical records, the system of record becomes a data and distribution position for cross-sell, retention, and risk signals, far beyond cost savings.

Customer-Specific Continuous Evals Imply a Compounding, Per-Customer Data Moat

Buried in a technical aside, Andrew (CEO) notes the system must improve models in a direction aligned "not just with... the way that servicing should be done generally but customer specific" [00:00:13]. If each servicer's workflows, regulations, and outcomes train an eval loop on the shared system of record, switching costs rise with every loan and every tuned agent. That makes Valon's platform stickier than a typical software vendor's, and explains why capturing the system of record first matters more than any AI feature layered on top.