Extracting data from Phase I ESAs
An environmental consultant writes a Phase I ESA for each property, following the ASTM practice in its own report template. Commercial real estate lenders and acquisition diligence teams receive one per site, dozens for a multi-site portfolio. Each arrives as a PDF, often hundreds of pages with appendices, through a data room or email.
Reviewers need the property and acreage, the component dates, and every recognized, controlled and historical environmental condition. A lender's environmental risk review confirms the report is still viable at closing, then decides whether the loan or purchase waits on Phase II sampling.
What makes Phase I environmental site assessments hard
Facts inside long narrative
The values a lender needs sit in prose across dozens of pages, such as a property's acreage given mid-sentence in the site description.
A clock set by components
Each component must fall within 180 days before acquisition, such as the site reconnaissance, which the report lists apart from its issue date.
Tables that break across pages
Regulatory database and nearby-site tables run over several pages, with a facility's name or notes split at the page break.
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 Phase I ESA, end to end
About this document: A Phase I ESA by Kimley-Horn for the City of San Jose, covering a 3.6-acre parking lot and taxiway strip at the city's airport, which posted it online. The file is the 39-page report body, without its appendices. It runs through site history, a regulatory database review over three pages of tables and the nearby release sites. Page 6, the executive summary, is featured.
Processing objective: The objective is to capture the engagement and its clock: who commissioned the assessment, for how much land, under which standard, and when its components were completed. The first three sit in the opening paragraph of running prose. The dates come from a summary table below it.
Demonstration fields: Client, Property size (acres), ASTM standard, Site reconnaissance date, EP declaration date
The document
4 of 39 pages
Page 1
Page 6shown below
Page 14
Page 21
Where each value came from
report.client
{
"value": "City of San Jose",
"page": 6,
"range": {
"end": 6698,
"start": 6581
},
"box": {
"xmin": 0.11374,
"ymin": 0.19021,
"xmax": 0.88838,
"ymax": 0.20665
}
}property.acreage
{
"value": 3.6,
"page": 6,
"range": {
"end": 6810,
"start": 6699
},
"box": {
"xmin": 0.11368,
"ymin": 0.20811,
"xmax": 0.88784,
"ymax": 0.22458
}
}report.astm_standard
{
"value": "E 1527-21",
"page": 6,
"range": {
"end": 7379,
"start": 7270
},
"box": {
"xmin": 0.11403,
"ymin": 0.29685,
"xmax": 0.88679,
"ymax": 0.31331
}
}component_dates.visual_inspection
{
"value": "May 12, 2025",
"page": 6,
"range": {
"end": 8961,
"start": 8949
},
"box": {
"xmin": 0.56625,
"ymin": 0.68747,
"xmax": 0.88282,
"ymax": 0.71549
}
}component_dates.ep_declaration
{
"value": "November 8, 2025",
"page": 6,
"range": {
"end": 9046,
"start": 9030
},
"box": {
"xmin": 0.56624,
"ymin": 0.71549,
"xmax": 0.88288,
"ymax": 0.74416
}
}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.
