Test whether your document can be extracted automatically.
Upload one invoice, bill of lading, commercial invoice or ACORD form. In about twenty seconds you see which fields an AI reads off it, how sure it is about each one, and what an automation would do with the data next.
Yes, most clean PDFs and scans of business documents can be read automatically. This test checks yours: it reads one document, lists every field it finds with a high, medium or low confidence flag, and flags anything a person should review. It is a test of one document, not a bulk converter.
This tool opens soon
The free tester is not switched on right now. In the meantime, the 14-Day Pilot runs this same kind of test on your real documents.
Upload one document
Reading your document. This takes about twenty seconds.
Here is what an automation can read from your document.
Fields found
| Field | Value | Confidence | Page |
|---|
A person should check
What automation would do next
Email me this and how you would automate it
Optional. We send the fields and a short plan for this document type. We never see the document itself.
Done. Check your inbox.
We will send the result and the plan shortly. Reply to that email with any question.
A feasibility test on your own document.
Most vendors show you a demo on their documents. This reads yours, and tells you where it is unsure, so you know before you buy anything whether the paperwork can be automated.
-
Upload one document
A PDF, PNG, JPG or WebP file, up to 10 MB and 5 pages. Pick the document type, or let the tool detect it.
-
See the fields it found
Each field comes back with the page it was read from and a confidence flag. Line items come back as a table. Anything uncertain is marked for review.
-
See what automation does next
The result ends with the steps a finished automation would take with this data: validate it, match it, route it, post it to your system.
Limits
- One document per test. Files up to 10 MB and 5 pages.
- PDF, PNG, JPG and WebP. Password-protected PDFs are refused.
- A few tests per visitor per day, and a daily cap across the site, so the tool stays free.
- Use a sample or a redacted copy. Do not upload Social Security numbers, bank details or medical records.
Privacy
- The file is read in memory and discarded as soon as the result is returned. It is never written to disk or kept.
- We keep only counts: the document type, the page count, how many fields were found, and whether the read worked. Never the content.
- The document is read by an AI model (Anthropic Claude) through its commercial API. It is not used to train anything.
- Your email is only collected if you ask us to send the result. Then it is used to reply and for nothing else.
See the privacy policy and security page.
What people ask about the document extraction test.
Anything else: hello@kwestra.com
Can AI extract data from my PDF or scanned document?
In most cases yes. Clean PDFs, photographed invoices and scanned forms can be read into labeled fields. Handwriting, heavy stamps, skewed phone photos and low-resolution scans lower the confidence. This test shows you which fields on your own document come out clearly and which need review.
Is the document extraction test really free?
Yes. You do not need an account or a credit card, and you see the result on the page without giving an email. Each visitor gets a few tests per day so the tool stays free for everyone.
Do you store my document?
No. The file is processed in memory and discarded when the result is returned. We keep only metadata such as the document type, page count and number of fields found, never the content.
What file types and sizes does it accept?
PDF, PNG, JPG and WebP, up to 10 MB and 5 pages. It reads one document per test.
How accurate is the extraction?
Accuracy depends on the document. Every field carries a confidence flag: high means the text was clear and unambiguous, medium means it was inferred or slightly unclear, low means a person should check it. Treat this as a fast test of feasibility, not a guarantee. A production automation adds validation rules and a review step for low-confidence fields.
Which documents work best?
Invoices, bills of lading, commercial invoices and ACORD insurance forms are the documents it is tuned for. Other business documents with printed fields, such as purchase orders and packing lists, usually work too: choose "Something else".
Can I use this to process a whole folder or inbox of documents?
No. This tool tests one document at a time. Running the same extraction on every document that lands in your inbox, with validation and posting to your system, is what the 14-Day Pilot scopes and prices.
What happens if I ask for the result by email?
We email you the extracted fields and a short note on how we would automate that document type for your team. A person from Kwestra may follow up once. You can reply to opt out.