Notes from the keyboard. Not the conference circuit.
Field reports from production deploys. Contrarian theses about what AI actually fixes in mortgage and home equity. Tactical playbooks teams can copy. Roughly one post a month.
Why 95% of AI projects fail — and why building in-house makes it worse
MIT found 95% of enterprise AI pilots show no P&L impact; RAND puts AI failure above 80%, twice the non-AI IT rate. The pattern isn't the model — it's the build, and in-house makes it worse.
Mortgage AI got the document work. The judgment is where it's stuck.
AI in mortgage handles document classification (63%) and reading (54%) but underwriting decisions sit at 21%. The judgment gap isn't a model problem — it's a trust-engineering one.
Real-estate agents adopted AI. They still won't trust it on a price.
82% of agents use AI (RPR) but 63% name output accuracy as their top concern. The gap isn't adoption anymore — it's trust, and trust is engineered, not prompted.
The AI-lite problem: why most mortgage vendor AI stops at the LOS door
AngelAi CEO Pavan Agarwal named the pattern: 'AI lite — safe but shallow.' Most lender AI sits in chat or lead nurture, kept away from the LOS. Vendors who can't integrate sell it as caution.
Home Equity Investment is a $4B asset class with no LOS
$4B+ in HEIs have been securitized in 2026, but every operator built their own stack. Hometap, Point, Splitero, Unlock, Unison — none of them bought one off the shelf.
Anatomy of an HEI eligibility engine: four investor overlays, zero LLMs in the decision
Field report from a multi-program HEI operator. Four investor overlays plus a sale-leaseback product, all driven by one admin-editable settings table. No LLM in the eligibility decision path.
You can't hire a senior AI engineer fast enough — here's the math
Senior AI engineers in 2026 cost $300K–$500K all-in and take six months to land. For a $2–10M revenue operator, that math doesn't work. Here's what does.
How we cut a home equity platform's lead-to-close cycle in half
Field report from a multi-program HEI / HEA / SLB platform. Voice intake, agentic enrichment, eligibility integration. What we shipped, what broke, what stuck.
The 7-specialist review pipeline that keeps AI features from breaking in production
A reusable playbook: seven structured review passes before any AI feature ships. Catches the failures that pilot-stage AI code carries into production.