Extract data from corporate tax returns
A corporate income tax return reports a full year of operations to the tax authority, and the version most US finance teams handle is IRS Form 1120 for a C corporation, or Form 1120-S for an S corporation that passes income through to its owners. A controller closing the fiscal year on 2025-12-31, a CFO reviewing the tax provision before signing, an outside CPA firm preparing several hundred returns before the 2026-04-15 deadline, and a lender underwriting a business loan against two years of filings all read the same document for the same short list of numbers: gross receipts, cost of goods sold, deductible operating expenses, taxable income, the federal tax computed on it, credits claimed, estimated payments already made, and the refund or balance due that falls out at the bottom. The figures resist a quick copy because each one rolls up from a schedule behind it. Gross income comes off a receipts schedule, cost of goods sold off a separate computation, and the deductions spread across wages, rents, depreciation, interest, and charitable contributions before they net into taxable income. The Employer Identification Number, the federal tax classification, and the principal business address tie the filing to one legal entity. Estimated quarterly payments made through the year offset the computed liability, and a credit such as the Section 41 research and development credit reduces it again. One number keyed to the wrong line changes the balance due and, at worst, the notice the company gets back from the IRS. Talonic reads the return and returns the identity block, the income and deduction lines, the tax computation, and the estimated-payment history as structured fields. Gross receipts of $8,400,000 net down to a taxable income of $1,350,000 and a computed tax of $283,500 that reconcile line by line, so a preparer or an underwriter starts from figures that tie out instead of retyping a scanned PDF.
What gets extracted from corporate tax returns
How extraction works for corporate tax returns
Corporate returns reach a preparer as e-file copies from tax software such as CCH Axcess, Lacerte, and UltraTax, as PDFs printed for the client file, and as scans pulled for a loan file, so the page layout shifts even though the form does not. Talonic classifies each return by form and tax year, then maps it to the corporate tax schema in the Field Registry, which binds each value to its line rather than its position on the page. Gross receipts, cost of goods sold, and the deduction schedule are read into their own fields, and the tax computation is checked so that taxable income at the corporate rate reconciles against the printed tax. Estimated quarterly payments are captured as a dated series and netted against the liability to derive the refund or balance due. Every figure returns with a confidence score and a pixel-region pointer under DIN SPEC 91491 conformity, so a controller or a lender can trace Line 30 taxable income or the balance due back to the source return.
Sample extraction
A 2025 IRS Form 1120 for a C corporation with an R&D credit
{
"form": "1120",
"tax_year": 2025,
"buyer.name": "Harbor Freight Components Inc.",
"ein": "12-3456789",
"federal_tax_classification": "C corporation",
"currency": "USD",
"gross_receipts": 8400000,
"cost_of_goods_sold": 3150000,
"gross_profit": 5250000,
"operating_expenses": 3900000,
"taxable_income": 1350000,
"tax_amount": 283500,
"tax_credits": 45000,
"total_tax_after_credits": 238500,
"estimated_tax_paid": 260000,
"tax_refund_or_balance": -21500,
"preparer.name": "Delta and Rowe CPAs LLP",
"estimated_quarterly_payments": [
{
"quarter": "Q1",
"payment_date": "2025-04-15",
"payment_amount": 65000
},
{
"quarter": "Q2",
"payment_date": "2025-06-16",
"payment_amount": 65000
},
{
"quarter": "Q3",
"payment_date": "2025-09-15",
"payment_amount": 65000
},
{
"quarter": "Q4",
"payment_date": "2025-12-15",
"payment_amount": 65000
}
]
}Frequently asked
Does it handle Form 1120 and the S corporation Form 1120-S?
Yes. Both are classified by form and tax year and mapped to the same corporate tax schema. The C corporation return carries an entity-level tax computed at the flat 21% rate, while the 1120-S passes income through to shareholders on Schedule K-1, so the pass-through figures are captured rather than an entity tax.
How is the tax computation checked?
Taxable income is reconciled against the income and deduction lines that produce it, and the computed tax is checked against taxable income at the applicable corporate rate. A return where the printed tax does not match the computation is flagged before it reaches a loan file or a provision workpaper.
Are estimated quarterly payments captured?
Each quarterly estimate is returned with its payment date and amount as a dated series, and the total is netted against the computed liability to derive the refund or balance due, which is the figure a lender and a controller both look for first.
Can it read attached schedules and depreciation detail?
When the deduction schedule, the cost-of-goods computation, and the depreciation and amortization detail are bundled in the same PDF, each is read into its own structured block and linked to the summary line it rolls into, so an analyst can drill from taxable income down to the wages or interest that reduced it.
Ready to extract from your own corporate tax returns?
Author note
Reviewed by Talonic engineering · last reviewed 2026-07-06