Key Moments
How Valon Is Rebuilding Mortgage Servicing for the AI Era
a16zKey Moments
Mortgage servicing, a $13 trillion market, is stuck in pre-internet tech. Valon rebuilt it from scratch, translating regulations into software and proving it with their own loans before selling it, achieving 3x efficiency.
Key Insights
The mortgage servicing industry operates on legacy systems built before the internet, managing a $13 trillion market.
Valon's approach involved becoming a regulated mortgage servicer, translating decades of federal and state regulations into software.
The company achieved 3x efficiency compared to traditional servicers, transforming break-even businesses into ones with 70-80% operating margins.
Valon's strategy to build from scratch was chosen over selling software to incumbents or acquiring existing servicers due to regulatory and IP challenges.
The ability to apply AI in a heavily regulated industry is identified as a key skill set for the next decade, with Valon's team learning by doing.
The core infrastructure Valon is building supports any system with money movements, regulation, and an operational component, not just mortgages.
The untapped potential of an archaic industry
The mortgage servicing industry is described as one of the most undisrupted sectors, still relying on technology from before the internet era. This represents a massive opportunity, with $13 trillion in consumer debt managed by legacy systems. Valon's mission is to build the foundational infrastructure for this market, enabling continuous learning and improvement of models aligned with both general servicing standards and customer-specific needs. This focus on critical infrastructure, which underpins any system involving money movement, regulation, and operations, is seen as a key area for innovation.
From stochastic calculus to systems engineering pain
Andrew Wang's initial interest in mortgage servicing stemmed from a fascination with the stochastic calculus involved in prepayment models. However, he quickly realized that practical application was more about estimation than rigorous math. This led to a deeper understanding of mortgage servicing as a complex systems engineering problem, characterized by vast amounts of data, significant money movement, and the constant need for reconciliation. His personal pain point of internal systems failing to reconcile money accurately at Soros propelled his desire to build better infrastructure and invest in companies that could achieve this.
The human cost of outdated systems
For the average homeowner who pays on time, the legacy systems might not be immediately noticeable. However, for those facing unexpected life events, the inadequacies of current systems become acutely stressful. A critical issue highlighted is the lack of historical context captured by legacy systems. For example, when a borrower passes away, their children trying to manage the estate might be denied access to past loan records, crucial for tasks like filing estate taxes. Traditional servicers often store such historical data in ad hoc databases, making it inaccessible if the current user isn't the original borrower. This lack of fidelity and contextual data causes significant anxiety and stress during already difficult times, underscoring the need for systems that can comprehensively track the life of a mortgage.
Addressing escrow payment challenges
Escrow accounts, which hold funds for taxes and insurance, present another challenge exacerbated by rising costs. Homeowners often struggle to make large, suddenly increased payments. Legacy systems are typically designed to calculate monthly payments based on an annual average, with limited flexibility to adjust quickly to new regulations or investor standards that allow for longer-term payment plans. This often requires an ad hoc approval process for modifications. Valon's approach aims to provide systems that can automate these decisions, allowing for immediate approval of modified payment plans, offering homeowners comfort and stability during financial strain, especially when unexpected events like natural disasters increase insurance premiums.
Choosing the hardest path: Building from scratch
Valon considered three paths: selling software to incumbents, acquiring and integrating existing servicers, or building a new servicing operation from the ground up. Selling software to incumbents was deemed ineffective due to heavy regulation and the risk of merely creating a prettier UI for outdated processes. Acquiring servicers presented intellectual property issues and forced platform design decisions. Therefore, Valon chose the most challenging route: establishing its own regulated mortgage servicer and building the software infrastructure from scratch. This allowed the software to evolve organically alongside the growing servicing operation, starting with a single loan and scaling to nearly a million.
Navigating the regulatory labyrinth
Securing licenses and translating decades of complex federal and state regulations into functional software was a monumental task. Key requirements, such as needing to be profitable before obtaining certain licenses, created chicken-and-egg problems. Valon strategized by initially focusing on non-QM loans that only required state licenses, carefully plotting the order of approvals. This involved navigating stringent state requirements, often needing government loan approvals or specific real estate broker endorsements, and demonstrating extensive experience across all servicing aspects like collections, customer service, and foreclosure. The process is described as extraordinarily long, often taking three to five years, requiring aggregation of business and expertise across numerous domains. Even getting approved for New York took three years, with the team having to physically track down an application in a mailroom to avoid resetting the clock.
Refactoring law into code: The COVID-19 catalyst
Translating regulations into software involved meticulously reading and refactoring every federal regulation (RESPA, TILA, FDCPA, etc.) and the mortgage, debt collection, foreclosure, privacy, and escrow regulations for all 50 states. This process created an abstract framework adaptable to diverse legal statutes. The COVID-19 pandemic proved to be a catalyst, with the team dedicating 18 hours a day for six months to reading, annotating, and developing schematics, a period characterized by 'pure grit, effort, pain, suffering,' followed by five to six years of testing.
Proving value through operational efficiency
Valon's initial strategy for attracting customers, primarily asset managers, was through superior unit economics and efficiency. By improving the technology, Valon could lower its operational costs and pass those savings on as lower prices, leveraging the economic sensitivity of asset managers. The company achieved approximately three times the efficiency of traditional servicers, transforming break-even businesses into operations with 70-80% operating margins. This efficiency gain was used to create market urgency and attract initial clients. The company's own servicing operation, managing around $200 billion in loans, served as a live proof of concept, undergoing rigorous state exams and audits, thus validating the safety and efficacy of their system before selling the software.
The shift from servicer to industry-wide platform
While Valon could have continued scaling its own servicing operation, the founders' original vision was to fundamentally change the industry. Selling the servicing operation to Carrington and then offering the Valon OS as a software platform to the industry allows them to address the core infrastructure problem rather than just keeping the 'alpha' for themselves. This aligns with their mission to be a generational company focused on fixing broken infrastructure, moving beyond just making money to solving significant societal problems, exemplified by their goal to 'save the world from mainframes.'
Generative AI: Orchestrating complex scenarios
The advent of functional generative AI has significantly expanded Valon's capabilities beyond what was imaginable at their Series A. AI now enables the orchestration of complex agent-led workflows, transforming how operators manage problems. For instance, during a disaster, instead of manual processes, AI can direct agents to make calls, offer various plans, and run specific workflows tailored to local conditions. This allows for greater customizability, enabling servicers to become more specialized and responsive, moving away from commoditized, cookie-cutter approaches. The ability to automate a higher percentage of tasks, including complex long-tail scenarios, and orchestrate a wide variety of responses, allows for improved portfolio outcomes through 'champion challenger' tactics.
Redefining human-system interaction with AI
Valon is rethinking how humans interact with systems, moving beyond traditional UIs and human operators skilled in navigating them. They have built harnesses on top of their system of record, allowing for classic chat interfaces for back-office operations. For homeowners, interaction can occur through voice AI or chat AI. This versatility in form factors opens up new possibilities for how individuals engage with financial systems, making them more accessible and intuitive.
Ensuring correctness and continuous learning in AI
A key technical challenge for Valon is ensuring correctness, particularly in a high-stakes industry like mortgages, where 'hallucinated' answers are unacceptable. Their approach involves aligning models to prioritize stating 'I don't know' over providing incorrect information. More broadly, Valon is building infrastructure for a continuous learning process. Unlike contained environments, servicing involves complex, dynamic ecosystems with accounting, money movements, customer interactions, and regulatory aspects. The system is designed to continuously run evaluations, pull data, and improve models in a direction aligned with both general servicing best practices and customer-specific needs, tackling a highly dynamic problem at scale.
Servicing as the foundation for broader expansion
The architecture and infrastructure developed for mortgage servicing position Valon for expansion into other regulated industries. Servicing is identified as the nexus of highly regulated enterprises. Beyond residential mortgages, this includes commercial real estate, where tenant financials are accessible, and healthcare, where revenue cycle management is essentially servicing for hospitals. Valon's deep operational and data insights gained from tackling the complex and 'sticky' mortgage market provide a strong foundation for addressing similar challenges in other sectors. The core concept of 'everything is servicing' suggests a vast potential for Valon's underlying infrastructure.
Proven traction and outcome-oriented partnerships
The market has validated Valon's approach, with the company signing over $200 million in deals within six months of launching as a software company, indicating significant pent-up demand for change. Their partnership with NewRes, where Valon sub-serviced their new loans, provided a clear demonstration of the software's performance. Valon structures customer engagements around outcome-oriented principles, aligning their earnings with their customers' revenue or cost savings. This core alignment was a key factor in securing large deals, such as the transfer of 4 million loans from Rhythm, signaling strong conviction from customers in Valon's capabilities.
A culture of high agency and impactful problem-solving
Valon attracts individuals who want to take initiative and drive impact. Many employees have prior experience or offers from top tech companies but choose Valon for the opportunity to solve a 'really important problem' and tangibly contribute to fixing broken infrastructure. The company fosters a culture where employees are surrounded by like-minded, driven individuals, leading to high retention rates, with a significant portion of the management team having been with the company for over five years. This demonstrates a successful culture built around people who thrive on tackling challenging, seemingly impossible problems.
Change management as the key to enterprise deployment
Deploying technology into large, complex, regulated enterprises is less a technology problem and more a change management challenge. Valon has spent significant effort developing its 'change management muscle' to navigate organizations, influence thousands of people with diverse objective functions, and drive transformation at scale. This understanding of organizational dynamics is identified as the number one value driver for the next decade, especially in the context of AI adoption. Thriving in these roles requires individuals with high agency, comfort with ambiguity, deep customer empathy, and the ability to bridge different perspectives to achieve a common goal.
Mentioned in This Episode
●Companies
●Organizations
Valon's Approach to Modernizing Mortgage Servicing
Practical takeaways from this episode
Do This
Avoid This
Valon's Efficiency Gains in Mortgage Servicing
Data extracted from this episode
| Metric | Traditional Servicer | Valon |
|---|---|---|
| Operating Margin | Break-even | 70-80% |
| Efficiency | 1x | 3x |
Common Questions
The mortgage servicing industry operates on legacy systems built before the internet, with many incumbents still using technology from the 1960s. This lack of modernization has created a massive opportunity for companies like Valon to innovate.
Topics
Mentioned in this video
A company rebuilding mortgage servicing for the AI era, focusing on technological innovation and regulatory compliance.
Mentioned as one of the regulatory bodies whose approvals are required for mortgage servicing licenses, and whose audits Valon has passed.
Referred to in the context of where talented individuals might typically work, contrasting with the unique impact-driven environment at Valon.
Mentioned as one of the regulatory bodies whose approvals are required for mortgage servicing licenses, and whose audits Valon has passed.
A partner of Valon that transferred 4 million loans to their system, validating Valon's performance and product-market fit.
The company to which Valon recently announced the sale of its mortgage servicing business, allowing Valon to focus on its software platform.
Mentioned as a competitor in the AI space where talented individuals might work, highlighting Valon's distinct focus on solving fundamental infrastructure problems.
Mentioned as an example of a company that used its platform to facilitate small business financing, analogous to how Valon could expand into commercial mortgage servicing.
Real Estate Settlement Procedures Act, one of the federal regulations Valon had to read and incorporate into its code.
Truth in Lending Act, one of the federal regulations Valon had to read and incorporate into its code.
Fair Debt Collection Practices Act, one of the federal regulations Valon had to read and incorporate into its code.
Telephone Consumer Protection Act, one of the federal regulations Valon had to read and incorporate into its code.
More from a16z Deep Dives
View all 64 summaries
25 minHow Microsoft is Adapting to the AI Era
22 minCybersecurity in the Agentic Era | Deep Dives with a16z
23 minDatadog CISO on Securing AI Agents at Scale | Deep Dives with a16z
48 minHow AI Is Redefining What It Means to Be Creative | Deep Dives with a16z
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free