Auto-build a site selection brief for every new property in Airtable

Build me an agent workflow that automatically writes a one-page site selection brief every time a new candidate property is added to my Airtable site pipeline.

Trigger: poll Airtable for a new record in my 'Site Pipeline' table. Each row carries a property address, a latitude, and a longitude. Skip any row that already has a value in the brief link field so we never produce duplicate documents.

For each qualifying new row, the agent should do the following:

1) Call Sites USA Request Trade Area Boundaries to compute 1, 3, and 5 minute drive-time polygons around the coordinate. Keep the WKT for each ring.

2) Call Sites USA Get Demographics by WKT for each ring and pull total population, median household income, age mix (for example shares of 18 to 34, 35 to 54, 55 plus), and daytime population. If a ring returns no data, flag that ring in the report as a data gap rather than failing the run.

3) Call Sites USA Get Merchant Counts by WKT for each ring to count competitors in the relevant retail category (the category should be configurable, but assume the user passes a category id or name).

4) Call Sites USA Get Store Visits by Shopping Center for the nearest shopping center to the candidate coordinate, and pull recent monthly visit counts so we can describe foot traffic context.

5) Create a Google Doc with Google Docs Create Document, titled with the property address, then use Batch Update Document to lay out a one-page brief with these sections: a short trade area summary, a demographic comparison table across the three rings, competitor density per ring, a foot traffic context paragraph for the nearest shopping center, and a recommended 'fit score' (0 to 100) with two or three sentences of reasoning. Note any data gaps inline.

6) Call Airtable Update Record on the originating row to write the Google Doc URL into the brief link field and the numeric fit score into the fit score field.

Other behavior: Sites USA enforces 1 call per second, so pace the calls. Treat any individual Sites USA call that returns no data as a flag in the report, not a fatal error. Do not retry the whole workflow on partial data. If the run fails before the doc is created, leave the row untouched so the next poll picks it up again.

Inputs the user should configure when setting this up: the Airtable base id and 'Site Pipeline' table id, the field names for address, latitude, longitude, brief link, and fit score, the retail category (id or name) used for competitor counts, and the list of demographic variables to include (default to total population, median household income, daytime population, and the age cohort breakdown).