05/13/2026
I think the CRE industry is about to learn a painful lesson:
An unknown number of multimillion-dollar deals have already been priced with AI-generated numbers no one can reproduce.
Not because people are careless.
Because the output looked clean.
That is the line I’m trying to draw in this piece:
AI is great for research, summarization, market color, and first-pass narrative.
But the numbers that price deals, size debt, and drive LP reporting need to be deterministic, traceable, and auditable.
If you can’t answer “where did this NOI assumption come from?” in one click, that is a real risk.
I’m not anti-AI.
I’m anti using language models for load-bearing math.
I wrote this after watching too many teams blur that boundary, buying into “AI black box” tools or prompt-only workflows for underwriting.
If you can’t trace a number to a source and reproduce it on demand, it doesn’t belong in a multimillion-dollar deal model.
AI is transforming how CRE professionals research, draft, and summarize. But underwriting, debt sizing, waterfall distributions, and LP reporting need engineered software — not generated paragraphs. Here's why.