HERE ARE LENDERS APPROVING MILLION-DOLLAR LOANS IN MINUTES WITH NO HUMAN INVOLVED.

There are lenders right now approving million-dollar loans in minutes. Datasets, valuation algorithms, instant

term sheets. No phone call. No site visit. No human being anywhere in the decision.

The pitch is speed and scale. And for standardized lending, consumer loans, conforming mortgages, automation genuinely works.

Construction and rehab lending is not standardized. I want to walk you through why the industry is splitting into two camps that are both wrong, and what we do instead.

1. The Full-Automation Mistake

An algorithm can calculate loan-to-value in a millisecond. Here is what it cannot do.

It cannot detect optimism in an ARV. It cannot walk a job site and notice the grade problem the budget ignores. It cannot call two contractors who worked with the borrower last year and hear the pause before they answer. It cannot know that permitting in one Valley suburb runs 3 weeks and in the next one over runs 4 months.

Someone once said more fiction has been written in Excel than in Word. I underwrite deals every week that look flawless on paper. The spreadsheet is precise. Precision is not accuracy. A model with confident inputs and wrong assumptions does not give you a wrong answer. It gives you a wrong answer you trust.

2. The Full-Manual Mistake

Here is where I break from the traditionalists: refusing the technology is not discipline. It is nostalgia.

AI and software do everything they are actually good at. Reading documents and flagging inconsistencies. Pulling market data in seconds that used to take an afternoon. Monitoring the portfolio. Processing draws so borrowers get funded in 24 hours instead of 3 weeks. I use AI as a sounding board on every major decision. It stress-tests my thinking and it never gets tired or emotional.

A lender doing everything by hand is slow, inconsistent, and drowning in operations. That is not safer. Slow underwriting is just error with worse service.

3. What We Do Instead

The machine verifies the math. Humans verify reality.

Every ARV on every loan is valued in-house. My wife Camille is a licensed agent here in Phoenix and she runs her own comps on every single deal. We do not rely on borrower ARVs. We do not rely on algorithmic estimates. Across every loan that exited in the first half of 2026, our internal valuations came within 2.3% of actual sale prices. That number is the whole argument.

Then there are the Four Cs: Character, Capacity, Collateral, Capital. Three of the four are human questions. I walk projects. I call references. I have been investing in this market since 2014, rentals, flips, syndications, which means I have been the borrower. I know what optimism looks like in a rehab budget because I have written optimistic rehab budgets. You cannot download that. You have to have paid for it.

4. The Gap the Big Lenders Cannot Close

Credit where it is due: the national lenders are getting genuinely good at the technology. Better than most people realize.

But their tech scales everything except the two things that decide outcomes in this asset class: local knowledge and relationships. They comp from a database. We comp from having stood in the kitchen. Their model saw a zip code. I saw the muddy lot with the drainage problem two streets off the arterial.

The machines are getting better everywhere. Judgment is getting scarcer everywhere. I am long judgment.

The takeaway, whether you are evaluating a lender, a fund, or any operator: ask what the machine does, what the humans do, and whether the person making the final call has ever operated on the asset they are underwriting. All three answers matter. Any private credit manager who cannot answer them crisply is asking you to trust the black box.

Where in your own investing are you trusting a model you have never pressure-tested? Reply to this email. I read every response.

Devon

P.S. Full automation is precision without accuracy. Full manual is conviction without verification. The edge is refusing to choose. Newsletter Edition #203