How Valon Rebuilt a $13 Trillion Industry From Scratch

Episode Summary

Executive Summary: The episode explores how Valen rebuilt mortgage servicing from scratch to modernize a $13T, heavily regulated industry still running on legacy systems. The founders explain why they became a servicer first, how they encoded complex regulations into software, and why operating the system themselves proved its value. They also detail how AI now enables far richer automation, orchestration, and continuous learning across servicing workflows.

Main Topics: Mortgage servicing as critical legacy infrastructure (Priority: 5/5): The speakers frame mortgage servicing as one of the largest, most outdated industries, with infrastructure predating the internet and massive room for modernization. Why Valen became a servicer before becoming software (Priority: 5/5): The company chose the hardest path—building and operating its own servicer—to avoid being constrained by incumbents and to prove the technology at scale. Regulation translated into code (Priority: 5/5): Valen had to read and encode federal and state servicing rules, build an abstract framework for compliance, and navigate licensing constraints across states and agencies. Operational pain points in legacy servicing (Priority: 4/5): Examples like borrower death, escrow recalculations, insurance changes, and payment hardship show how poor systems create stress for homeowners and agents. AI’s new role in servicing workflows (Priority: 5/5): Generative AI now enables agents, voice/chat interfaces, complex workflow orchestration, and continuous evaluation loops that were previously impractical. Commercial expansion beyond mortgages (Priority: 4/5): The team argues servicing is a universal infrastructure layer across regulated industries, including healthcare revenue cycle management and commercial real estate. Culture, deployment, and change management (Priority: 4/5): Large enterprise deployments require high-agency, high-empathy teams and strong change-management skills, not just good technology.

Key Arguments: Mortgage servicing is a giant, under-modernized market with $13T of consumer debt still supported by legacy infrastructure. Traditional software-only entry was not viable because regulated servicers would not adopt a de novo system and would constrain product design to legacy expectations. Building and operating Valen as a servicer first provided proof, safety, and urgency for selling the platform later. Encoding regulation into software required translating federal and state legal requirements into an abstract, testable system rather than relying on ad hoc processes. Customer pain often comes from faulty infrastructure, not frontline service; better UX alone cannot fix incorrect system-of-record data. AI changes the product from deterministic automation to orchestrating complex, dynamic workflows with agents, escalations, and localized playbooks. The platform’s value comes from continuous learning: running evals, improving model behavior, and aligning outcomes with customer-specific needs. Servicing is a reusable infrastructure pattern across industries where money movement, regulation, and operations intersect.

Data Points: Mortgage servicing market size: $13 trillion - Used to describe the scale of consumer debt still managed by outdated systems. Legacy technology age: Pre-internet / 1960s-era architecture - The incumbents’ systems were described as being built before the internet and architected in the 1960s. Efficiency improvement: ~3x as efficient - Valen said its technology made the business roughly three times more efficient. Operating margin: 70%–80% - Estimated margin level after turning a break-even servicing business into a software-enabled business. Initial software revenue: $200 million+ in deals - Signed within six months of going to market as a software company. Largest servicing transfer: 4 million loans - Rhythm’s transfer to Valen was described as the largest servicing transfer. Share of market: Almost 10% - The 4 million-loan transfer was said to represent nearly 10% of the market. Scale of growth: From 1 loan to close to 1 million - The servicer grew from a single loan at launch to nearly a million loans. Series A deck length: ~60-page PDF - Referenced as the original detailed plan for the company’s build-to-software strategy. COVID regulation work cadence: 18 hours/day for 6 months - Andrew described intensive regulation-reading and system design during COVID. Licensing timeline: 3–5 years - Typical minimum forecast for obtaining the necessary approvals and licenses. New York approval time: 3 years - They received New York approval in record time, though they still never got licensed to originate there. Management team tenure: 75%–80% - Approximate share of management team with five-plus years at the company.

Pivotal Quotes: "Mortgage is probably top three in terms of undisrupted industries that exist." — Linda: Opening framing of the market opportunity and its outdated state. "The skill set that people will be hiring for in the next decade is the ability to apply AI." — Linda: Explaining why building AI into the hardest regulated workflows matters now. "Everything is actually servicing." — Andrew: Describing servicing as a universal infrastructure layer across regulated, operational businesses.

Implications: The episode suggests regulated industries will be reshaped by companies that own both operations and software. AI’s biggest near-term impact may be in complex workflows requiring compliance, context, and continuous learning—not just generic automation.