Extract data from financial statements
Unstract reads balance sheets, income statements and cash flow statements, audited, reviewed or scanned, and returns checked figures to your spreading template, loan system or decision engine.
| 2025 | 2024 | |
|---|---|---|
| ASSETS | ||
| Cash and cash equivalents | 412,880 | 298,410 |
| Accounts receivable, net | 1,284,310 | 1,102,775 |
| Inventories | 966,120 | 1,018,440 |
| Prepaid expenses | 58,200 | 61,900 |
| Total current assets | 2,721,510 | 2,481,525 |
| Property and equipment, net | 3,108,640 | 2,874,300 |
| Total assets | 5,830,150 | 5,355,825 |
| LIABILITIES AND STOCKHOLDERS' EQUITY | ||
| Line of credit | 600,000 | 450,000 |
| Current maturities of long-term debt | 318,400 | 302,100 |
| Accounts payable | 742,980 | 688,215 |
| Total current liabilities | 1,661,380 | 1,440,315 |
| Long-term debt, less current maturities | 1,946,300 | 2,264,700 |
| Stockholders' equity | 2,222,470 | 1,650,810 |
| Total liabilities and equity | 5,830,150 | 5,355,825 |
need FY23 + 1120-S to tie — JL
See accompanying notes and independent accountant's review report.
{ "company": "Harlow Precision Machining", "statement": "balance sheet", "period_end": "December 31, 2025", "basis": "reviewed", "total_assets": 5830150, "total_current_liabilities": 1661380, "stockholders_equity": 2222470, "current_ratio": 1.64, "confidence": 0.96}
Trusted by finance teams around the globe
Every accountant lays out a balance sheet differently
A clean PDF straight from the audit firm is the easy case. Most borrower packages look like this.
Scanned from a bound annual report, every page crooked
Three years, two accountants, three layouts
Company-prepared figures, corrected in pen by the owner
Balance sheet, income statement, cash flow and notes in one PDF
Trial balances and debt schedules exported to spreadsheets
An analyst keys every line into the spreading template before credit can look at the deal, and one slip carries into every ratio.
Unstract takes that on — it reads statements in the accountant's own layout, pulls the lines you spread, checks that they tie and match the tax return, and sends them to your spreading template, loan system or decision engine.
Parse, extract and spread statements on one platform
Read the statement, turn it into figures, run the spread around it. Each product works alone, and they stack when you need the whole process.
Parse financial statements with every line and year in place
Capture the full text of a balance sheet, income statement or cash flow statement, whether it came as a clean PDF, a scan of a bound report or a fax from the borrower, with its columns intact. Each figure stays on its line and under its year, so 2024's receivables are never read as 2025's, and a bracketed loss keeps its sign.
Send the text to your own LLM, or make parsing the first step of a spreading workflow.
Learn more about LLMWhispererTurn financial statements into structured figures that tie out
Name the lines you spread, from cash and receivables down to net income and capital expenditures, and every statement comes back in the same JSON structure, whatever the accountant called each line. A scan or a company-prepared statement gives you the same record. Each figure is checked as it comes out: a total that doesn't foot, or net income that disagrees with the tax return, arrives flagged with the reason.
Call it as an API from your loan system, run it as an ETL pipeline from a shared folder, or make it one step in an orchestrated credit process.
Learn more about Document StudioRun the credit spread from package to decision
Describe your spreading process in plain English and Flow's agents build the workflow. The package arrives from the relationship manager or the portal, every statement, tax return and bank statement is read, and the figures go into your spreading template the same way for every borrower. DSCR, leverage and liquidity are worked out and checked against your credit policy, and the spread goes to your decision engine or loan system, such as FICO Platform, nCino or Moody's CreditLens. You approve the workflow once, and it runs identically on every package.
Set the conditions for a person to step in: a policy exception, statements that don't reconcile, a missing year. Only those packages stop, and the credit analyst sees why.
Learn more about FlowExtract the figures now. Automate the spread when you're ready.
Start for free lets you parse and extract your own statements today. Automating credit spreading starts with a consultation.
The full statement package, and the reports around it
Documents
- Balance sheets
- Income statements
- Cash flow statements
- Statements of stockholders' equity
- Notes to the financial statements
- Audited, reviewed and compiled statements
- Company-prepared statements
- Interim and year-to-date statements
- Trial balances
- General ledger reports
- Operating statements
- Accounts payable aging reports
- Annual reports
- Business income proof
- Debt schedules
- Personal financial statements of guarantors
Fields captured
For the teams that spread and review financials
Borrower statements spread the same way every time, so analysts review the credit instead of keying the numbers.
Quarterly statements from every portfolio company read on arrival and checked against covenants.
Client trial balances and statements turned into workpaper-ready figures, whichever software produced them.
Subsidiary and acquisition-target statements in one structure, ready to consolidate or compare.
Faster spreads, fewer keying errors
Safe for a borrower's confidential financials
A statement package shows a borrower's revenue, margins, debt and owner distributions, often with their tax returns attached.
- Encrypted in transit and at rest.
- Borrower documents are never used for model training.
- Access by role, team and portfolio, and every edit to a spread recorded against the package.
- You set retention: statements are deleted on the schedule your policy requires.
- Self-hosted deployment for lenders whose credit policy keeps borrower data inside their own environment.
What lenders and analysts ask before they start
Templates map positions on a page, so they break when an accountant changes the layout. Here each line is read by what it is: "Trade receivables, net of allowance" and "Accounts receivable" both land in the same field, so a new borrower's statements need no new template.
LLMWhisperer reads scans, faxes and photographed pages with the columns intact, so each figure stays under the right year. Document Studio then returns the figures as JSON through an API endpoint per document type.
Unstract does it in three layers: LLMWhisperer parses the statement, Document Studio extracts the lines you name, and Flow runs the spread, checks and handoff around them. Use the first two on their own, or all three for the whole process.
Describe the values once, as you would to an analyst: cash, current maturities, EBITDA add-backs. Document Studio finds them under whatever label and in whatever order the accountant used, across every year on the page.
The whole package. Balance sheets, income statements, cash flow statements and the notes are read together, alongside the tax returns, bank statements and debt schedule that arrive with them, so figures can be checked across documents.
Yes. A 120-page annual report is split into its statements, notes and schedules first, and each is read on its own. You get the figures from the primary statements and the notes you asked for, not the chairman's letter.
Use AI for the reading and keep the analysis deterministic. Document Studio uses LLMs to extract the figures; ratios such as DSCR, leverage and liquidity are then calculated by rules you approve in Flow, the same way every time, and the credit decision stays with your policy and your analysts.
Yes. Subtotals are footed, the balance sheet is checked to balance, net income is compared with the tax return and the change in equity, and debt is compared with the debt schedule. Anything that doesn't tie is flagged with both figures.
Yes. You define the schema to match your spreading template or your loan system's import, and every statement comes back in that shape, ready to load without remapping.
Yes. Document Studio deploys an endpoint per document type: POST a statement package and get the spread-ready figures back as JSON. The same extraction runs as an ETL pipeline if you prefer dropping files in a folder or bucket.
Unstract is SOC 2, ISO 27001, GDPR and HIPAA compliant. Statements are encrypted in transit and at rest, never used to train models, and kept only as long as you set. Lenders that need data to stay in-house can self-host; see the Trust Center.
In a consultation we set it up around your borrowers' statements, your spreading template and your credit policy, and most teams spread their first live packages within days. To try the extraction now, start for free and upload a balance sheet.
Bring three borrowers' statements to the consultation
Three years of statements, the tax returns and a debt schedule, scanned copies included. We spread them live and show every line, every check and each exception.