Extracting data from W-2 forms
Every US employer issues a W-2 to every employee each January, on a layout the IRS fixes box for box. What differs is the rendering, and every payroll vendor has its own. Tax preparers and mortgage underwriters take them by the thousand, as portal downloads and phone photographs of the mailed copy.
A preparer needs the wages in box 1, the federal tax withheld in box 2, and the state and local rows along the bottom. Tax software carries those onto a return, and a loan origination system reads box 1 to qualify a borrower's income.
What makes W-2 forms hard
Same boxes, no two alike
The IRS fixes every box and its number, and each payroll vendor renders that layout its own way on its own stock.
Box 12 codes carry meaning
A single letter beside each box 12 amount says what the money is: D marks a 401(k) deferral, DD the cost of health coverage.
Marks and write-ins among numbers
Box 13 is three checkboxes and box 14 is an unconstrained write-in, sitting on a form that is otherwise printed amounts.
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 W-2 form, end to end
About this document: An example Form W-2, Copy A, published as an illustration on the Remote People blog. One landscape page, supplied as an image with no text layer, so each value is read from pixels. All of it is obvious test data: employee John Doe, employer Company Inc, and a locality named Test.
Processing objective: The objective is to read one of each kind of value the form carries. That means a printed amount, a ticked checkbox, a free-text write-in, and a figure from the state and local row. Each grounds to its own box, label and value together.
Demonstration fields: Wages (box 1), Social security wages (box 3), Third-party sick pay checked (box 13), Other (box 14 write-in), Local income tax (box 19)
The document
1 page
Page 1shown below
Where each value came from
federal.wages_tips_other_compensation
{
"value": 2323,
"page": 1,
"range": {
"end": 261,
"start": 220
},
"box": {
"xmin": 0.53868,
"ymin": 0.08766,
"xmax": 0.75533,
"ymax": 0.15043
}
}federal.social_security_wages
{
"value": 2323,
"page": 1,
"range": {
"end": 485,
"start": 454
},
"box": {
"xmin": 0.53865,
"ymin": 0.1504,
"xmax": 0.75532,
"ymax": 0.21315
}
}box_13.third_party_sick_pay
{
"value": true,
"page": 1,
"range": {
"end": 1190,
"start": 1166
},
"box": {
"xmin": 0.68762,
"ymin": 0.46317,
"xmax": 0.75529,
"ymax": 0.526
}
}box_14_other
{
"value": "Other Income $100",
"page": 1,
"range": {
"end": 1290,
"start": 1261
},
"box": {
"xmin": 0.53888,
"ymin": 0.52598,
"xmax": 0.75523,
"ymax": 0.68283
}
}state_and_local.local_income_tax
{
"value": 675,
"page": 1,
"range": {
"end": 1646,
"start": 1617
},
"box": {
"xmin": 0.74268,
"ymin": 0.68309,
"xmax": 0.88274,
"ymax": 0.74518
}
}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.
