Rent Comp Survey

Why I built it

A weekly rent comp survey means getting the lowest and highest advertised rent, square footage, concessions and fees for every floorplan at every competing community. Done by hand it takes someone most of a day per market. Owners and lenders ask for it and our leasing teams price from it, so it has to be right and it has to be on time. I wrote down how I run it with AI and published it.

What it does

• The method, the workbook template, paste-ready prompts, return contracts, and checking rules.
• Single-agent run for small markets; multi-bot pipeline (intake, one bot per property, checker) for twenty-plus communities.
• Helper scripts that clone last week's tab, parse return blocks, diff against the prior week, and print a flag report.
• Automatic distribution of the finished workbook to ShareFile, Dropbox, Drive or a webhook.
• A full fictional example run, from reference file to check report.

How it works

One property per context window. I learned this the hard way. A bot holding three communities in its head will file a floorplan under the wrong one, and the mistake looks exactly like real data. Each community gets researched on its own and written to disk before the next one starts.

The research bots never see last week. A separate checker diffs this week against last, assembles the notes and flags what a person should look at.

Nobody estimates. If a rent isn't published, the sheet says so and carries last week's number forward with a note.

Calls I made