Five hundred listings ran out of a spreadsheet. Now the search site, the CRM and the owner's own AI operator run on one database the agency owns.
A public search site, a CRM for leads, viewings and deals, a photo pipeline, a private AI operator and daily reporting, all on one Postgres database. A web enquiry and an agent's note land in the same place.
Search by keyword, reference code, type, district, beds, price and size, with 48 named BTS and MRT stations as filters, a clickable Bangkok map, buying-cost calculator and neighbourhood pages. Old WordPress URLs redirect in.
live at propickproperty.com since 23 June 2026, four days after the rebuild call, about 8 weeks for the full build
For buyers and renters, the agency's own teamrobots for 12 named AI crawlers, sitemap, llms.txt, listing schema, canonical metadata, shipped with the site.
16 clicks in launch month June 2026 to 90 clicks and 5,516 impressions by mid September 2026, average position 22 to under 11 (Search Console)
every enquiry stores its source, medium, campaign and referrer; AI assistant referrers are tagged.
the agency knows where a lead came from, including ChatGPT
a morning script diffs CRM rows, GA4 sessions and Search Console clicks and flags what changed.
a daily view of changes across the system
Numbers are dated database and Search Console readings. Client revenue and deal values are never published. June was the launch month, so part of the search growth is a new site getting indexed.
Share the problem or the brief. We’ll review it with you and propose a clear first scope.
Discuss a similar project Explore more work