WorkflowsWorkflows

How to Build a Property Operating Budget with AI

A full-year operating budget that normally takes days can be assembled from a property's source documents in 1-2 hours using Claude, with every number source-cited.

A Denver apartment building rendered as a line drawing feeding source documents into an operating budget spreadsheet
Illustration: AI for CRE

A full-year operating budget that typically takes days of manual work can be assembled from a property’s source documents in 1-2 hours using Claude. The method was demonstrated on a 2027 budget for an apartment property in Denver, drawing from a single deal folder of roughly 15 files: rent roll, operating statement, debt summary, investment summary, market comps, and a staffing roster. The same approach applies to any property.

Have the model build the prompt first

The workflow starts before any budget is written. Instead of requesting the budget directly, the model is asked to draft the prompt and pose clarifying questions first, so no assumptions get built into the numbers. Claude reviews the folder, then returns a short set of questions that each determine how the budget is calculated:

The model is also instructed to work in the role of the asset manager. Answering these questions up front locks the logic before any calculation runs.

Choosing a model

The demonstration used Sonnet 5, which handles a budget of this scope. Opus is the stronger option for heavier Excel work. Either produces the result.

Building the budget

With the prompt set, the full deal folder is attached. Claude confirms all 15 files are present, then works through the budget sheet by sheet, a process that took roughly 15 minutes given the volume of documents.

The output is a four-tab Excel workbook:

Each figure carries its source in plain language. Payroll and benefits are budgeted at a 13.2% increase built from the staffing roster at the position level; utilities are set at a 20% increase. Because every line documents where it came from, the budget can be traced and defended without a separate review.

Turning the process into a reusable skill

The final step converts the one-time workflow into a repeatable skill that runs the same way on every deal. The model again asks setup questions: whether to create a new skill or replace an existing one, whether to assume fixed file names (no, since each deal folder differs), whether to lock the four-tab output (yes), and where files should live (one folder per deal).

A skill defined only by an instruction file will not reliably reproduce the same output. The working version, named Deal Budget Builder, includes three components: the instruction file, a references folder holding a document classification guide, a QC checklist, and the completed Denver budget as a worked example, and a template folder containing the budget as a blank Excel file. With the approved template stored inside the skill, every future budget returns in the same structure, with the same tabs, columns, and basis notes.

The result is a repeatable budgeting process that produces source-cited output on any property directly from its existing deal documents.

We did a full demo video inside the AI for CRE Collective here: https://www.skool.com/ai-for-cre-collective/about

AI for CRE Collective