Extracting data from loan histories
A servicer keeps a line for every payment a loan has taken, printed on whatever report its software produces. A buyer reads the history before the loan trades. A court reads it when a borrower disputes what was paid, and servicing teams produce thousands a month, arriving as PDF exports.
A reviewer needs each payment's due and received dates, what it was split into and the balance it left behind. A boarding or audit system replays the history row by row and flags the payments that arrived late or went nowhere near principal.
What makes loan histories hard
One table, dozens of rows
A loan's whole life runs as thirteen ruled columns over several landscape pages, with the header repeated at each break.
The irregular rows are the point
A partial payment, a late one and a short one sit among the regular rows, and those are the rows a reviewer is looking for.
Blank cells beside filled ones
A reference column is empty on one row, zeroes on the next and a number on the one after, with nothing to mark the difference.
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 loan history, end to end
About this document: A loan payment history produced by Loan Servicing Soft, published as a sample of its software's output. Two landscape pages: a header with the servicer, loan and borrower, then one ruled table of 57 payments in thirteen columns, carried over the page break with its header repeated, and a total row. Page 1 is featured. The data is fictitious.
Processing objective: The objective is to lift one payment that went wrong, out of a table where every row looks alike. It fell due on the first of the month and arrived on the twentieth. It carried a late fee, and nothing at all reached principal.
Demonstration fields: Date due, Date received, Late fee, Applied to principal, Principal balance
The document
2 of 2 pages
Page 1shown below
Page 2
Where each value came from
payments[7].date_due
Printed on the page as 09/01/2010, normalized on extraction.
{
"value": "2010-09-01",
"page": 1,
"range": {
"end": 1897,
"start": 1887
},
"box": {
"xmin": 0.01901,
"ymin": 0.34933,
"xmax": 0.10346,
"ymax": 0.37385
}
}payments[7].date_received
Printed on the page as 09/20/2010, normalized on extraction.
{
"value": "2010-09-20",
"page": 1,
"range": {
"end": 1916,
"start": 1906
},
"box": {
"xmin": 0.10346,
"ymin": 0.34935,
"xmax": 0.18731,
"ymax": 0.37386
}
}payments[7].late_fee
{
"value": 22.25,
"page": 1,
"range": {
"end": 2029,
"start": 2024
},
"box": {
"xmin": 0.7133,
"ymin": 0.34944,
"xmax": 0.77899,
"ymax": 0.37393
}
}payments[7].principal
{
"value": 0,
"page": 1,
"range": {
"end": 1985,
"start": 1981
},
"box": {
"xmin": 0.50494,
"ymin": 0.34941,
"xmax": 0.58252,
"ymax": 0.3739
}
}payments[7].principal_balance
{
"value": 49864.47,
"page": 1,
"range": {
"end": 1957,
"start": 1949
},
"box": {
"xmin": 0.31663,
"ymin": 0.34938,
"xmax": 0.42242,
"ymax": 0.37389
}
}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.
