Pam Golding Properties agents were spending 45+ minutes per listing writing property descriptions — a repetitive task requiring them to study property photos, interpret condition and style, and write copy that matched the agency's brand standards. With 60+ listings per month, this consumed roughly 45 hours of agent time monthly on copy alone.
The client needed something that fit seamlessly into their existing workflow (the Property Book system), not a separate tool they'd have to adopt.
I built a Google Chrome extension that injects directly into the agents' existing Property Book interface. No new tab, no separate login, no workflow change — just a new button on the page they already use.
The extension scrapes property details (type, size, location, features, price) and image URLs from the current listing page using Chrome Extension APIs.
OpenAI Vision API analyzes each property photo — determining materials (timber floors, granite counters), condition (pristine, well-maintained), style (contemporary, colonial), and notable features (fireplace, wrap-around veranda).
Based on property type, price bracket, and condition signals, the system retrieves the appropriate writing guidelines from a curated library — developed collaboratively with the agency's marketing team to match their exact brand voice.
Claude generates the property description using the scraped data, Vision analysis, and selected guidelines. Output is structured: headline, opening paragraph, feature highlights, closing.
An automated QA pass checks for brand voice compliance, completeness, and accuracy. The system also generates 3–5 suggested edits for the agent to review, focusing on personalization and local market context.
Average description time dropped from 45 minutes to 9 minutes — an 80% reduction. Agents still spend time reviewing and refining, but the heavy lifting is done. Quality improved because Vision AI catches details human writers miss when quickly scanning photos.
Brand consistency improved significantly. Descriptions now reliably match agency tone across all agents and property types, reducing the editing burden on senior staff.
The system processed 60+ listings per month in production, saving the agency roughly 36 hours of agent time monthly.