A document-to-listing pipeline that ingests developer project and unit documents (brochures, floor-plan PDFs, price sheets, spec packets) and extracts structured fields — project name, location, home types, unit mix, pricing, completion timelines, and amenities — onto a clean listing data model
Real EstateDevelopers send the documents; the AI writes the listing, prepares the images, and publishes it live — no manual content, design, or data entry.
Real Estate / PropTech — new-construction presale & assignment marketplace (British Columbia, Canada)
listingsnearby.com
Workflow
- Intake
- Extract
- Generate
- Review
- Publish
project.pdf
fields.json
listing ✓
Human approval gate before publish
The challenge
Listings Nearby is an online marketplace for new-construction presale and assignment homes across British Columbia (Surrey, Vancouver, Burnaby, Langley, Abbotsford). Developers hand over project and unit information as raw documents — brochures, floor-plan PDFs, price sheets, spec packets — and turning each one into a polished, published listing meant slow manual content writing, image preparation, and data entry. Every new project was a bottleneck, and the catalog could only grow as fast as a human could retype it.
Real Estate
Real Estate / PropTech — new-construction presale & assignment marketplace (British Columbia, Canada)
What we built
The system, in parts.
AI listing-content generation that turns the extracted data into ready-to-publish copy: project descriptions, unit summaries, and search-friendly detail pages, on-brand and consistent across the catalog
AI image handling that prepares and assigns project and unit visuals to each listing, so a developer's media becomes a complete, presentable listing without manual design work
Auto-publish to the live marketplace: validated listings go straight onto the map-based and city-based search experience — no manual content writing, image work, or data entry in the loop
Human-in-control guardrails: extracted and generated listings are reviewable before they go live, so the team stays in control of accuracy and presentation while the AI does the heavy lifting
Outcomes
What changed for them.
A developer's raw project documents become a live, structured listing with written content and imagery — collapsing days of manual content and data-entry work into an automated flow
The marketplace catalog can scale with new developer projects without scaling a content or data-entry team behind it
Listing content and imagery stay consistent across the whole site instead of varying with whoever typed up each project
New presale and assignment projects reach buyers faster — less lag between a developer sharing documents and the listing being searchable
The team keeps oversight: every auto-generated listing is reviewable before publish, keeping accuracy and brand control with humans
How it’s built
The stack.
Engagement summarized from delivery records; some figures are directional.
Document extraction (PDF/brochure parsing)LLM-based content generationAI image processing & assignmentStructured listing data modelAuto-publish pipeline to live marketplaceMap-based / city-based searchHuman-in-the-loop review
More work
See allTwo ways to start
Want a system like this?
Tell us the workflow you want to run itself. We will scope a focused first project — designed, built, and operated, with humans in control.