Extracting data from audited financial statements
A CPA firm audits an organization's annual statements and signs an opinion in front of them. Layouts vary by audit firm, and a nonprofit's balance sheet ends in net assets. Lenders collect one from each borrower every year, and buyers request several years of them in diligence, as emailed PDFs or data room downloads.
A credit analyst needs the opinion type, total assets, total liabilities, operating cash flow and the year's net income or change in net assets. Spreading software loads them into the lender's financial template, which tests covenant ratios against the loan agreement.
What makes audited financial statements hard
The opinion is written in prose
The opinion type, accounting framework and report date sit inside the auditor's letter, set in paragraphs beneath a letterhead and above the firm's signature.
Each entity type reports differently
A company's balance sheet closes on shareholders' equity, a nonprofit's on net assets split by donor restriction, a government's on net position.
Two years side by side
Comparative statements print the current and prior year in adjacent columns of similar figures, with the year-end date printed once above each column.
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 audited financial statement, end to end
About this document: The audited financial statements of Family First, Inc., a Florida nonprofit, as posted publicly by the Florida Auditor General. Its 27 pages hold the independent auditor's report, the four basic statements, the notes, and a Single Audit section of award schedules and compliance reports. Page 7 is the statement of financial position, featured here. Page 9 is a landscape statement of functional expenses.
Processing objective: The objective is to show the balance sheet closing. The assets side builds from cash to a total, and the liabilities side builds to a matching total that includes net assets. Both land on $15,689,274, boxed on two different lines of the same page.
Demonstration fields: Cash and cash equivalents, Total current assets, Total assets, Total liabilities, Total liabilities and net assets
The document
4 of 27 pages
Page 4
Page 7shown below
Page 9
Page 20
Where each value came from
statement_of_financial_position.cash_and_cash_equivalents
{
"value": 14021341,
"page": 7,
"range": {
"end": 8861,
"start": 8849
},
"box": {
"xmin": 0.54405,
"ymin": 0.27425,
"xmax": 0.90509,
"ymax": 0.29341
}
}statement_of_financial_position.total_current_assets
{
"value": 15252147,
"page": 7,
"range": {
"end": 9112,
"start": 9102
},
"box": {
"xmin": 0.54361,
"ymin": 0.34793,
"xmax": 0.90525,
"ymax": 0.3746
}
}statement_of_financial_position.total_assets
{
"value": 15689274,
"page": 7,
"range": {
"end": 9325,
"start": 9313
},
"box": {
"xmin": 0.54317,
"ymin": 0.4286,
"xmax": 0.90542,
"ymax": 0.45472
}
}statement_of_financial_position.total_liabilities
{
"value": 1758660,
"page": 7,
"range": {
"end": 9799,
"start": 9790
},
"box": {
"xmin": 0.77372,
"ymin": 0.66307,
"xmax": 0.90712,
"ymax": 0.69052
}
}statement_of_financial_position.total_liabilities_and_net_assets
{
"value": 15689274,
"page": 7,
"range": {
"end": 9946,
"start": 9934
},
"box": {
"xmin": 0.77377,
"ymin": 0.72551,
"xmax": 0.90709,
"ymax": 0.75115
}
}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. VPC and on-prem deployment require an Enterprise plan.
