ChatGPT Just Entered the Mortgage Market
ChatGPT just entered the mortgage market.
On April 21, Better launched Tinman — a full loan origination and underwriting engine — inside ChatGPT Enterprise. Loan officers can now process, underwrite, and fulfill loans conversationally. The system was trained on $110 billion in funded loans and 5 billion pages of credit documentation.
This isn't a demo. It's live.
And it changes one question for every mortgage lender: if your competitors can underwrite a loan conversationally in minutes — what does your verification workflow look like by comparison?
Here's what most coverage is missing:
AI agents that automate origination are only as trustworthy as the data they consume. Tinman can process a loan instantly. It cannot verify that the borrower exists, that the pay stub is real, or that the income figure wasn't generated by the same AI tools now being used to commit mortgage fraud at scale.
Speed without verification is just faster fraud.
The mortgage industry is building an AI-powered future at remarkable speed. The lenders who get this right will be the ones who establish deterministic, source-level verification at the data layer — before deploying probabilistic automation on top of it.
That's not a reason to slow down on AI. It's a reason to get the foundation right first.
What's your take — is conversational underwriting inside ChatGPT a breakthrough, or is the industry moving faster than the verification infrastructure can support?
— Stephen Schrump, CEO, PitchPoint Solutions
Ready to Transform Your Verification Process?
See how industry leaders are streamlining verification with PitchPoint.
Continue Reading
More insights you might find valuable
AI Fraud in Mortgage Lending: Why Rules Beat Guardrails
AI-generated fraud is accelerating—synthetic identities, deepfake pay stubs, fabricated bank statements. If your fraud detection is AI-based, it can be fooled by AI. Here's why source-level, deterministic verification is the only durable defense.
Stephen Schrump
May 13, 2026
65 of 100 Loans Fall Out — And You're Paying for the Bad Ones Twice
65 of 100 applications now fall out — and most lenders pay for the bad ones twice. The $1,610 hiding in every dead file, and the five checks your credit report will never run.
65 Out of 100 Loans Now Fall Out. The Real Cost Is Hiding in Your P&L.
In 2020, 65 of 100 applications closed. Today it's 35 — and most lenders are measuring the cost wrong. The true cost of fallout, the two-bucket framework, and how to move verification upstream before you spend a dime.
Stephen Schrump
June 29, 2026