Lesson 00 · Free · Open now

Inside the apartment model

The calculation blocks inside an apartment acquisition underwriting model, what goes wrong in each — and the seven ways a workbook an AI populated fails silently. Taught on a real model, with the real numbers on screen.

13 min 47 sec·No signup· Worksheet below

01 · What it covers

Seven calculation blocks, and the quiet lever inside each

01The rent rollIn-place against market rent — and the blank cell in a named row that corrupts six monthly engines without erroring.

02Expenses + renovationSign conventions, and the capex plan that carries the entire value-add thesis.

03NOI aggregationWhere every upstream error concentrates — which is exactly why NOI always looks clean.

04The pricing blockFour ways to set one price, and the single dropdown that decides which of them the model actually reads.

05The debt blockDSCR at the events the loan cares about — post-development, lease-up, refinance — and which constraint binds.

06Property returnsUnlevered against levered, and the basis-points field that turns a sensible input into a hundred-times error.

07The waterfallPref, promote hurdles and the split — the least-read cells in the file, where the most money moves.

02 · The heart of it

Seven ways an AI-populated workbook fails silently

Handing a model to AI is genuinely good — it is fast, it covers ground, and it leaves a written explanation. The question is not whether to use it. The question is what breaks when it breaks, because it does not break loudly.

The stale output

Formulas that never recalculated, reported fluently as if they had.

The dead input

A price written into a cell nothing downstream reads.

Target versus achieved

You asked for 8.50%. The file landed at 8.28%. The summary still says 8.50%.

The half-filled row

One blank cell. Outputs don't error — they move.

Units

Basis points or percent. Plausible either way, wrong by a factor of a hundred.

Structure zeroed

Hurdles out of order, or an equity share that silently voids every LP return.

Plausible is not sourced

A reasonable number with a beautiful justification and no source document behind it.

03 · Keep this

The worksheet

Two sides, A4. The calc-block map with every quiet lever named, and the silent-failure checklist to run before you trust any AI-populated workbook. Print it, pin it beside the screen.

Open the worksheet (PDF)

04 · The model in the lesson

Get the model and the skill

The lesson is taught on the A.CRE Apartment Acquisition Model, built by the team at Adventures in CRE — the models a generation of analysts learned on. Each one now ships with an AI skill through AI.Edge, their AI education platform. Both are free, and this lesson stands on their work.

aiedge.ac/join →

Where does your firm actually stand?

The Underwriting Integrity Index scores fifteen dimensions — how much AI is really in your workflow, and whether a number it touched can still be proven. Four minutes for the quick read, or a live interviewer that scores as you talk.

Run the Index →

05 · What comes next

Lesson 00 is the free one. The program is the discipline.

Eight lessons for CRE deal teams — from document extraction with a review discipline that holds, through verification gates, to letting AI work inside models with proof at the gate. Every exercise runs on one of your own live deals, and the courseware installs as working skills and role agents for your AI tools.

See the program → Talk it through first