What about footnoted values like 'see note 5'?

We extract the cell content as-is including footnote markers. Some downstream processing may want to strip them — toggle 'flatten footnotes' in Options.

Can I extract from multiple filings at once?

Yes — use the API's batch endpoint. Drop in 30 10-Ks; get 30 XLSXs out, named by ticker.

What if the table spans pages?

Detected automatically. The output sheet has continuous rows; the page break in the source is invisible in the output.

Use case · 5 minutes per filing

How to extract every table from a financial PDF into one spreadsheet

10-K, 10-Q, annual report → clean Excel with one sheet per table, ready to chart.

For: Equity analysts, investors, FP&A teams pulling data out of regulatory filings

Start with step 1 Browse all tools

STEP BY STEP

The workflow

Run AI Table Extract on the filing

Drop in the 10-K. AI Table Extract identifies every table — balance sheet, income statement, cash flow, notes — and detects column boundaries even when the source uses spacing-based pseudo-tables (common in old filings).

Open AI Table Extract

Verify column headers

Multi-row headers ('FY 2024', '2023', '2022') sometimes come through misaligned. Spot-check the headers on the financial statements you'll actually use; the model is right ~95% of the time, that 5% will bite you.

Open AI Table Extract

Export as XLSX with one sheet per table

Each detected table becomes a sheet, named after the heading from the filing. Charts you build against the sheets stay tied to the source.

Open AI Table Extract

WHY IT MATTERS

The case for doing this

Manually retyping financial-statement data is the worst kind of busy work. It's error-prone, slow, and the result is a one-shot deliverable that breaks if the filing is corrected. AI table extraction inverts the cost: 5 minutes of model work, instantly comparable across years and companies. For analysts covering 30 names, that's 30 hours per filing season recovered.

COMMON MISTAKES

What to avoid

Trusting numbers without verification

Models occasionally substitute digits — 0/O, 1/l, 5/S in particular. Sum the extracted column and compare to the printed total before relying on the data.

Extracting from low-DPI scans

Old SEC filings can be image-only at low resolution. OCR first; extraction quality is bounded by the OCR layer it's reading from.

Mixing currency formats

$1,234.56 vs €1.234,56 — the comma/decimal swap matters. Set the locale in the Options panel before extracting.

TIPS FOR THE BEST RESULTS

What works

Crop to the table area before extraction

Surrounding paragraphs sometimes confuse column detection. Crop the page to just the table for cleanest results.

Use AI Table Extract over PDF-to-Excel for messy tables

The standard PDF-to-Excel works on clean grid tables. Multi-row headers, merged cells, and footnoted cells need AI Table Extract for clean output.

Validate against printed totals

Every reputable filing has subtotal and total rows. If your extracted column doesn't sum to the printed total, you have an extraction error — usually a missed row.

RELATED WORKFLOWS

Other jobs people do with pdfcraft ai

How to combine bank statements into one PDF for your accountant

Stitch 12 monthly statements into one searchable, OCR'd PDF in under 5 minutes.

How to translate an employee handbook into 5 languages with consistent layout

One source PDF → 5 language-specific PDFs, layout preserved, terminology consistent.

How to turn a 50-page research paper into clean study notes

Hierarchical bullets, key definitions, equations highlighted — your own personal cliff notes in 2 minutes.

FAQ

Questions about this workflow

On clean digitally-generated PDFs (the modern norm for 10-Ks): 98%+ on numeric values, 95%+ on multi-row headers. On scanned filings, accuracy is bounded by OCR quality.

Try the workflow on a real PDF

Free tools run in your browser. No signup, no watermarks.

Start with step 1