Using AI: CRE Investor Edition
Using AI: CRE Investor Edition
Meet my Quant
Do you remember the scene from The Big Short, where Jared Vennett introduces his quantitative analyst, the math specialist who accurately calculated that an 8% default rate would result in MBS Armageddon?
Today, I’d like to introduce my personal quant, a trusty (for the most part) sidekick that makes my job A LOT easier. I’ll show you how I use chatGPT to quickly run numbers on CRE offerings, from basic metrics to more complex stress tests.
After reading this article, you will be able to upload a deck, have AI run various metrics (YoC, debt yield, DSCR with and without interest-only period), and create a variety of sensitivity tables (for example, exit cap rates to investor-level IRR)
I have bad news: the tool is only useful to those who possess the knowledge to interpret the results. Want to learn more? Start here:
How to Read a Real Estate Pro Forma series: Part 1, Part 2, Part 3, Part 4 and Part 5.
Before we dive in, let me remind you that this information is for education purposes only, and is not investment nor legal advice - do your own due diligence.
Garbage in - Garbage out
Let me tell you a story. Many moons ago, when dinosaurs roamed the Earth, I took a statistics class. The professor insisted we do all calculations by hand - even though back in those dark ages we had TI-84 calculators that could do all of it faster and better.
In the decades since, I have never manually plotted a bell curve again, but I’m forever grateful for that class. Why? Because I can easily spot errors when my TI-84 outputs something atrocious.
Think of AI as TI-84 on steroids: unless you have a general understanding of what you are looking at, you won’t be able to spot errors, let alone fix them.
The most important consideration: you can’t trust the output if the inputs are faulty. As they say, garbage in - garbage out. ALWAYS start by making sure your AI assistant of choice understands the task at hand and identifies the correct data to work with.
To run the following tests, I created a dummy offering presentation which I uploaded to chatGPT. All outputs were double-checked with the Excel spreadsheet that was used to underwrite the deal.
I start all my queries like this:
Prompt: Summarize the investor terms (waterfall, fees), capital structure, debt terms and expected returns
Any errors can be easily corrected at this stage.
Run the Basics
Before we get to more complicated stress tests, start here and check gpt’s work along the way:
Prompt: Calculate Yield on Cost, and present in a table:
Here’s why this metric is important:
