Key Takeaway: A cost-plus system where two million livelihoods depend on the inefficiency persisting.
Two million people make their living off the inefficiency. Loan officers, processors, underwriters, closers, title agents, appraisers, servicing specialists, default managers, foreclosure attorneys, REO brokers. Each role exists because the process is fragmented enough to require a human at every handoff. Each human adds cost. Each cost gets passed to the borrower. Everyone knows this. Nobody says it out loud. I spent five years building technology to fix pieces of this pipeline, and the thing I learned that no pitch deck ever captures is that the people inside the system aren't villains. They're rational actors responding to incentives that were baked into the infrastructure before they were born.
We just didn't think change was possible. Not because the technology didn't exist -- it did. Because the incentive structure was so entrenched, so distributed across so many participants, that nobody with the power to simplify it had the incentive to do so. The complexity was the product.
Cost-plus as business model
I cover the full cost structure -- eight layers of double marginalization, $15,000-$25,000 in stacked fees per closing -- in The Mortgage Value Chain and One Pipeline Pretending to Be Ten Businesses. The short version: every intermediary prices as "cost of inputs plus my fee," nobody sees the full stack, and the complexity each layer preserves is the complexity each layer profits from.
Why we thought it couldn't change
The barriers felt permanent: regulatory capture cementing intermediary structures through RESPA, TILA, and Dodd-Frank; 3,600 counties with incompatible record systems and no standard API; and institutional inertia where the people closest to the problem were the same people whose careers depended on it. The mortgage executive who sees the inefficiency also has a $400K salary that depends on it. The title company CEO who knows title insurance is a data infrastructure tax also knows their business model requires the tax to persist. But infrastructure problems are engineering problems.
What actually changed
AI and data engineering didn't change the incentives. They changed the minimum viable team to challenge them. A three-person pod can now build what used to require a 50-person company. The cost of attempting disruption dropped by an order of magnitude.
At Staircase, the strongest products were built by engineers with zero prior mortgage knowledge. They approached problems in the abstract -- pure disaggregation, uncontaminated by domain assumptions. Having too much domain knowledge biased people toward replicating existing broken processes. This was the Bezos insight applied to a startup: every team communicates through well-defined service interfaces. No back doors. Teams move independently.
The cost-plus structure will compress. Not because someone will disrupt it with a startup -- but because the data layer underneath it will become commoditized, and when the data is free, the intermediaries built on data scarcity lose their pricing power. Found money is better than printed money.
Why it still takes 45 days
Friction is the product. Nobody built a seamless mortgage process because that would let you shop around. The byzantine data freshness rules — different agencies requiring data no older than 2 weeks, 45 days, 60 days — force last-minute scrambling by design. If mortgage switching were as easy as switching cell carriers, the entire commission structure would collapse. So everyone in the chain has an incentive to keep the process painful enough that you never consider alternatives.