Extracting data from pay stubs
A pay stub accompanies every paycheck, and state law fixes what must appear on it without saying how. California names gross wages, hours, deductions and net pay; the layout is the payroll system's own. Mortgage underwriters and benefits agencies read them in the thousands, as portal PDFs and photographs of paper.
An underwriter needs gross pay, the tax and deduction lines, net pay, and the pay period dates. A loan origination system annualizes gross pay into qualifying income, and a benefits system tests net pay against an eligibility threshold.
What makes pay stubs hard
Two columns of every number
Each line carries the current period beside the year to date, so every amount appears twice with a different figure.
Codes without a key
Taxes and deductions are abbreviated to Med, SS, Federal and 401K, and each payroll system abbreviates them its own way.
Half a dozen small tables
Earnings, taxes, deductions, direct deposit and totals each get their own block and header row on a single page.
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 pay stub, end to end
About this document: A Lenovo direct deposit earnings statement for a single pay period. One portrait page, set largely in Courier, with six ruled tables and a footnote about business expenses. This is a real employee's statement. The name, address, employee ID and last four of the social security number were replaced with obviously fake equivalents before publication. The pay amounts are printed as they were.
Processing objective: The objective is to pull one paycheck's arithmetic off the page so it can be checked. Gross pay of $5,361.60, less $1,387.30 in taxes and a $268.08 deduction, leaves the $3,706.22 that was deposited. The Medicare line is included because 1.45% of gross pay is exactly the amount shown.
Demonstration fields: Gross pay, Medicare tax, Total taxes, 401(k) deduction, Net pay
The document
1 page
Page 1shown below
Where each value came from
totals.gross_pay
{
"value": 5361.6,
"page": 1,
"range": {
"end": 1162,
"start": 1153
},
"box": {
"xmin": 0.4054,
"ymin": 0.46594,
"xmax": 0.52627,
"ymax": 0.48244
}
}taxes[0].amount
{
"value": 77.74,
"page": 1,
"range": {
"end": 730,
"start": 724
},
"box": {
"xmin": 0.62941,
"ymin": 0.29966,
"xmax": 0.76654,
"ymax": 0.31484
}
}totals.total_taxes
{
"value": 1387.3,
"page": 1,
"range": {
"end": 1194,
"start": 1185
},
"box": {
"xmin": 0.62932,
"ymin": 0.46597,
"xmax": 0.76655,
"ymax": 0.48245
}
}deductions[0].amount
{
"value": 268.08,
"page": 1,
"range": {
"end": 1450,
"start": 1443
},
"box": {
"xmin": 0.25784,
"ymin": 0.52008,
"xmax": 0.38058,
"ymax": 0.70302
}
}totals.net_pay
{
"value": 3706.22,
"page": 1,
"range": {
"end": 1898,
"start": 1889
},
"box": {
"xmin": 0.44755,
"ymin": 0.8395,
"xmax": 0.64034,
"ymax": 0.8604
}
}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.
