Investing & underwriting

Build a hold, sell, or refinance decision model from a seven-file deal folder with one prompt, and compare the workbooks Claude, ChatGPT for Excel, and Shortcut produce

Jake Heller of PLOT Daily gave Claude, ChatGPT for Excel, and Shortcut, an AI agent built for Excel, the same prompt and the same deal folder for Meridian at Highline, a 196-unit apartment property with a loan maturing in 16 months. The folder held seven files: a rent roll, a T-12, a debt summary, an investment summary, a market comp summary, a capital improvements history, and the current budget. The prompt asked for a complete hold, sell, refinance decision model with formulas everywhere and no hardcoded calculations, built as six sheets in order: Assumptions, Hold analysis, Sale analysis, Refi analysis, Summary dashboard, and Recommendation matrix. Shortcut asked four questions first (projection horizon, GP/LP waterfall, refinance terms, dashboard detail), recommended a hold-period input cell, wrote a plan, and built a workbook with blue input cells and a sensitivity analysis. Claude asked an unneeded question and paused whenever focus left the workbook; ChatGPT for Excel started immediately and hit script errors. All three recommended holding, but the returns differed materially, and Shortcut's sell-now scenario showed a loss that looked wrong against the debt summary. The article's conclusion: the reusable asset is the prompt, and a person traces each workbook through the Assumptions sheet before the model reaches a partner or lender. The formulas were not audited cell by cell.

At a glance

Tools used
ClaudeChatGPTExcelShortcut
Category
Investing & underwriting
Audience
investor
Difficulty
Copy-paste
Source
We Tested Claude, ChatGPT, and Shortcut on the Same Hold-Sell-Refi Model
Jake Heller (PLOT Daily / AI for CRE Collective) · Article · Aug 10, 2026
First step
Put the rent roll, T-12, debt summary, investment summary, market comp summary, capital improvements history, and current budget for one property with an upcoming loan maturity into one folder.

Compliance note: Three tools given one prompt and one deal folder produced materially different returns, so at least two of the workbooks were wrong somewhere. Trace each output through the Assumptions sheet and check the loan payoff and refinance sizing before a hold, sell, or refinance model reaches a partner, investor, or lender.

Suggested prompt

Written by this directory, inspired by the source. Replace the [bracketed] parts.

Using the property data in this workbook and the attached deal folder (rent roll, T-12, debt summary, investment summary, market comp summary, capital improvements history, and current budget), build a complete hold, sell, refinance decision model. Use formulas everywhere and no hardcoded calculations, so every figure traces back to the Assumptions sheet. Build these sheets in order: 1) Assumptions, 2) Hold analysis, 3) Sale analysis, 4) Refi analysis, 5) Summary dashboard, 6) Recommendation matrix. Before you build, list any decision the documents cannot answer, such as projection horizon, GP/LP waterfall terms, and refinance terms, and ask me. Make the hold period an input cell. Then show me your plan before you write to the workbook.

Share on X Read the source

← All use cases