Extracting data from property inspection reports
Licensed home inspectors write a property inspection report after walking a house, each in their own report software with their own category labels. Buyers, agents and repair estimating platforms receive one for nearly every home under contract. A busy brokerage sees hundreds a month, as emailed PDFs or web report links.
A repair request needs each deficiency: its item number, the system it sits in, the inspector's category and its location on the house. A repair estimating tool prices each item, and the agent builds the repair addendum sent to the seller.
What makes property inspection reports hard
Findings are written as prose
Most of a report is numbered findings written as sentences, often opening with a category label and closing with a location on the house.
Labels vary by inspector
Report software and inspectors each choose their own categories, so one report's SAFETY CONCERN is another's Major Defect or Repair Now.
Summary cells hold several findings
The summary table groups deficiencies by system, so one cell for the crawlspace or the cooling system can stack several numbered findings together.
Parse, extract, verify
Parse
Parse turns the document into reading-order Markdown and a hierarchy of blocks (text, tables, figures), and grounds each block to its exact location on the page.
Extract
You define the fields in a JSON schema, and Extract returns their values in a consistent, structured format, each linked back to where it appears in the Markdown.
Verify
Each value links back to where it appears, down to the character range in the Markdown, and the page and bounding box in the source document, so every result can be verified.
One property inspection report, end to end
About this document: A residential inspection report on a two-story single family house, published by Home Inspector Pro as a sample of its report software. It runs to 30 portrait pages: site details, one section per system with numbered findings and photographs, and a two-page summary of significant deficiencies. Page 4, the roof and attic section, is featured. The client and address are specimen entries, so the file is shown as published.
Processing objective: The objective is to turn the roof section into structured findings a repair request can use. Each finding keeps its item number, the inspector's category label and where on the roof it was seen. Each value grounds to the passage of the report that states it.
Demonstration fields: Roof covering material, Recommendation category (item 2.2), Recommendation category (item 2.3), Damaged shingles item, Damaged shingles location
The document
4 of 30 pages
Page 4shown below
Page 10
Page 20
Page 29
Where each value came from
roof.covering_material
{
"value": "fiberglass based asphalt shingles",
"page": 4,
"range": {
"end": 8950,
"start": 8804
},
"box": {
"xmin": 0.05654,
"ymin": 0.21492,
"xmax": 0.9436,
"ymax": 0.4065
}
}roof_findings[1].category
{
"value": "IMPROVE",
"page": 4,
"range": {
"end": 9700,
"start": 9574
},
"box": {
"xmin": 0.06604,
"ymin": 0.4512,
"xmax": 0.91208,
"ymax": 0.46406
}
}roof_findings[2].category
{
"value": "MAINTENANCE",
"page": 4,
"range": {
"end": 9944,
"start": 9822
},
"box": {
"xmin": 0.06609,
"ymin": 0.4764,
"xmax": 0.91695,
"ymax": 0.48958
}
}roof_findings[3].item_number
{
"value": "2.4",
"page": 4,
"range": {
"end": 10187,
"start": 10061
},
"box": {
"xmin": 0.06632,
"ymin": 0.50169,
"xmax": 0.9135,
"ymax": 0.51443
}
}roof_findings[3].location
{
"value": "front upper left side",
"page": 4,
"range": {
"end": 10261,
"start": 10188
},
"box": {
"xmin": 0.06678,
"ymin": 0.5142,
"xmax": 0.56811,
"ymax": 0.52716
}
}How to try it yourself
Everything on this page is reproducible. The sample, its schema, and the Parse and Extract API outputs are published. You can repeat the same two calls from whichever surface you already build on.
- Claude Skilldescribe the job; the agent writes it
- CLIno code at all
- MCPexpose ADE docs as an agent tool
- REST APIany language
- Pythonlandingai-ade
- TypeScriptlandingai-ade
Documents into data
Accurate
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Auditable
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Agentic
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
API-first
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Affordable
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Enterprise-grade
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Built for high-volume, high-throughput production
- Zero data retention
- Priority rate limits
- Enhanced support
- VPC and on-prem deployment
Available on Team and Enterprise plans. VPC and on-prem deployment require an Enterprise plan.
