AI DOCUMENT FILLING

Forms that
fill themselves.

Vellum reads your reference documents and fills any blank form — PDFs, DOCX, scans — with cited answers. Drop the file in, get the filled file out.

Try the dashboard See how it works
Every format.
PDF, DOCX, scans, photos. Filled output preserves the original file.
PDF · DOCX · JPG · PNG · HEIC
Cited answers.
Every field links to the page and bounding box of the source document.
PAGE + BBOX · CONFIDENCE
Human in the loop.
Low-confidence fields are flagged for review before the form is signed off.
THRESHOLD < 0.70 → REVIEW
3 reference documents
PDF
lease_2026.pdf
12 pages · 2.4 MB
PDF
rental_application.pdf
4 pages · 840 KB
DOCX
house_rules.docx
3 pages · 120 KB
LEASE_2026.PDF
Reading 3 documents…
FILLING
Tenant name
—
Property address
—
Lease start
—
Monthly rent
—
Security deposit
—
Pet policy
—
How it works

Two steps. The blank form goes in, the filled form comes out.

01 · EXTRACT

Turn a form into structured questions.

Upload a blank form plus your reference documents. Vellum reads everything and pulls out every question, with cited answers and a confidence score.

INBlank form + referencesOUTStructured answers
02 · FILL

Drop the answers back into the form.

Vellum writes each answer into the correct field — checkboxes, dropdowns, signatures, scanned forms — and returns the document in its original file format.

INBlank form + answersOUTFilled form
Built for

Teams that fill the same forms, every week.

Legal
34 templates
Discovery, intake, NDAs
Pull party names and dates from contracts to populate a discovery response.
HR
12 templates
I-9, W-4, onboarding
Fill onboarding packets from a single offer letter and ID upload.
Finance
28 templates
KYC, audits, tax filings
Pull audit answers from a quarter of board minutes and policy docs.
Healthcare
19 templates
Prior auth, intake
Build prior-authorization requests from a chart and the patient’s history.
What it handles

The messy stuff,
handled.

Real-world forms aren’t tidy. They’re scans, handwritten notes, government PDFs from 2003. Vellum reads what’s actually there — typed text, checkboxes, signatures, tables — and writes answers back into the original document, pixel-aligned.

Scanned PDFs.
Photographs, faxes, and screenshots of forms. We OCR them, find every field, and write back to the original pixels.
Checkboxes & dropdowns.
Not just text inputs. Radio groups, multi-select boxes, and form-specific selectors are filled correctly.
Tables & line items.
Multi-row tables (expense lines, signers, dependents) are matched up by header and filled row-by-row.
Citations on every field.
Each answer points to the page and bounding box of the reference document it came from. No hallucinations, no guesses.
FAQ

Common questions.

PDFs (typed and scanned), DOCX, DOC, RTF, JPG, PNG, HEIC for forms. References can be any of those plus XLSX, CSV, and TXT. Output matches the input form’s format.

See it on your
own forms.

Open the dashboard
vellumdocs.app© 2026 VELLUM, INC.