Document data extraction
Deepen AI's trained teams extract fields from invoices, forms, receipts and contracts, or check and correct your AI's extraction, field by field.
From $5 per person-hour A plan within 2 working days No sales call
What a checked document looks like
Each document comes back approved, corrected or sent to you, with the reason for every change.
- Document
- Supplier invoice, 1 page
- Your AI extracted
Total: $2,145.00. VAT: $0.- Checked as
- CorrectedTotal: $2,145.00. VAT: $357.50.
- Reason
- On this supplier's layout the VAT line sits below the total, and the extraction missed it.
- Document
- Photo of a taxi receipt
- Your AI extracted
- Date: 14 March 2026. Amount: €38.20.
- Checked as
- ApprovedEvery field matches the receipt.
- Reason
- The date was read day-first, which matches your rules for European receipts.
- Document
- Handwritten insurance claim form
- Your AI extracted
Policy number: 88412-B- Checked as
- CorrectedPolicy number: 88412-8
- Reason
- The last character is an 8. Your policy numbers never contain letters.
- Document
- Lease agreement, 14 pages
- Your AI extracted
- Renewal notice: 60 days
- Checked as
- Sent to youClause 9 says 60 days, but the signed amendment on page 14 says 90. Which should be recorded?
- Reason
- Your rules don't cover amendments. Your answer is added to the guidelines.
How we start on your work
No sales call. A plan in your inbox, then a short paid pilot on your real documents.
What we need from you
- Your field list and rules, or a filled-in example
- 20 to 30 documents with the correct values
- Access to the documents, and where results go: your tool, a sheet or a bucket
- 5 minutesTell us the task
Which documents, how many, and how fast you need the fields.
- Within 2 working daysGet a plan and a price
People, hours, turnaround, start date and the total price.
- 2 weeksRun the pilot
We train the team on your fields and rules, process real documents and send two weekly reports. Then you decide.
Who extracts, and how accuracy is kept
Trained people, working in English inside your tools. Each one learns your fields and rules and passes a test on known documents before live work.
Every reviewer's accuracy is measured
Items you've already answered are hidden in the real work, so we know how accurate each person is.
Nobody guesses on unclear cases
The team lead decides, or asks you. Your answer joins the guidelines.
You see the results every week
A weekly report shows work done, QA results and what's improving.
Standard extraction covers invoices, receipts, forms, orders and contracts in English, typed or handwritten. Other languages and specialist documents are quoted in your plan.
How much does document data extraction cost?
$5per person-hour
Standard extraction in English. You pay for the hours people work, with a trained team, a team lead, QA checks and a weekly report included.
Set before you pay. Your plan shows the team size, hours and total price.
Starts with a two-week pilot on your real documents, paid upfront, with two weekly reports.
Your documents stay in your systems. We work under NDA with the access you set. Deepen AI holds SOC 2 Type II, ISO 27001, ISO 9001 and TISAX.
Questions
Can you check our AI's extraction rather than start from scratch?
Yes. People check each field your AI extracted against the document and correct what's wrong, which is faster than typing every field.
What about handwriting and poor scans?
People read handwriting and low-quality scans every day. Fields that can't be read with confidence are flagged for you, never guessed.
How is accuracy measured?
Field by field. Documents with known correct values are mixed into the work, QA re-checks a sample, and the weekly report shows the acceptance rate for each field.
How fast are documents processed?
You set the turnaround you need when we plan the work. We staff the team to meet it, and it's written into your plan, with the price, before you pay anything.
What format do results come in?
Wherever your process expects them: in your tool, in the columns of a shared sheet, or as JSON or CSV files in a bucket.
Our documents contain personal data. How is it protected?
People work inside your systems with accounts you control, under NDA, and you can remove access at any time. Data moves only if we agree it in writing first.