Scraped listings in, qualified sellers out.

Demo data is generated in the browser. In production the pipeline runs on self-hosted n8n: scrape, dedupe against the CRM data table, flag numbers on the Do-Not-Call registry, then Claude scores each owner's motivation signals — days on market, price cuts, absentee ownership, pre-foreclosure notices, tax arrears — and only score ≥ 7 triggers outreach drafted in my voice.

Same qualification brain, pointed the other way: the buyer's criteria are weighed against the scraped inventory — price fit, type, area, features, days on market — and each property gets a match score. High matches trigger a tour invite; mid matches go out as a personalized listing digest; the rest stay in nurture until inventory changes.