Extracting data from Form 1120 returns
Every C corporation files a Form 1120 each year, printed by its preparer's tax software on that year's IRS layout with schedules attached. Commercial and SBA lenders collect two or three years of returns from each corporate borrower, as PDFs uploaded to a portal or sent by the accountant.
Credit analysts need total assets, total income, taxable income, total tax and the balance sheet. Spreading software maps those into the lender's standard statement, where cash flow and leverage are calculated for the credit memo.
What makes Form 1120 returns hard
A balance sheet in four columns
Schedule L reports beginning and end of year in paired columns, so a single asset line can carry two amounts in each period.
Page 1 rests on schedules
Several page 1 lines carry totals from schedules and attached forms, such as dividends from Schedule C and the tax from Schedule J.
One amount on several lines
A return often repeats a figure down the ladder, such as taxable income before and after special deductions, each on its own numbered line.
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 Form 1120, end to end
About this document: The six pages of Form 1120 from an IRS e-file test scenario, published so tax software vendors can check their output. An invented corporation with page 1's income and tax lines, then Schedules C, J, K, L, M-1 and M-2. Page 1 is featured. Every value is IRS test data, and the officer and preparer signature lines are blank.
Processing objective: The objective is to run the return from top to bottom line. How large the corporation is comes first, then what it earned and deducted, and finally the tax owed against what it paid. Each figure grounds to its own amount cell on page 1.
Demonstration fields: Total assets (item D), Total income (line 11), Total deductions (line 27), Total tax (line 31), Total payments and credits (line 33)
The document
4 of 6 pages
Page 1shown below
Page 3
Page 4
Page 6
Where each value came from
corporation.total_assets
{
"value": 87608959,
"page": 1,
"range": {
"end": 1091,
"start": 1045
},
"box": {
"xmin": 0.74294,
"ymin": 0.15173,
"xmax": 0.94539,
"ymax": 0.18129
}
}income.total_income
{
"value": 77254243,
"page": 1,
"range": {
"end": 2747,
"start": 2737
},
"box": {
"xmin": 0.8236,
"ymin": 0.37878,
"xmax": 0.94284,
"ymax": 0.39374
}
}deductions.total_deductions
{
"value": 40415384,
"page": 1,
"range": {
"end": 4653,
"start": 4643
},
"box": {
"xmin": 0.8236,
"ymin": 0.62116,
"xmax": 0.94295,
"ymax": 0.63637
}
}tax_and_payments.total_tax
{
"value": 7719804,
"page": 1,
"range": {
"end": 5497,
"start": 5488
},
"box": {
"xmin": 0.82361,
"ymin": 0.71203,
"xmax": 0.943,
"ymax": 0.72746
}
}tax_and_payments.total_payments_and_credits
{
"value": 9000000,
"page": 1,
"range": {
"end": 5731,
"start": 5722
},
"box": {
"xmin": 0.82361,
"ymin": 0.74256,
"xmax": 0.94301,
"ymax": 0.7576
}
}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.
