Document AI for your own document types

Classify and extract any document.

Upload a few samples, confirm the fields DocWise suggests, and every future document of that kind is classified, extracted, validated and delivered to your systems — on the AI provider you choose.

Bring your own document types, or start from ready-made ones
Choose the AI provider your data goes to
Review only the fields the model is unsure about

Start free Read the API docs →

Example API response (illustrative data)

{
  "status": "COMPLETED",
  "document_type": { "name": "Supplier invoice" },
  "result": {
    "data": {
      "invoice_number": "INV-20931",
      "issue_date": "2026-09-30",
      "total": 1284.5,
      "line_items": [
        { "description": "Pallet wrap", "amount": 284.5 },
        { "description": "Freight", "amount": 1000 }
      ]
    },
    "fields": [
      { "path": "/total", "confidence": 0.98,
        "evidence": { "page": 1, "quote": "Total due 1,284.50" } }
    ]
  }
}

How it works

From samples to an automated pipeline without writing templates or prompts.

1
Upload a few samples
Drop three to five examples of a document you receive — invoices from your suppliers, a claims form, a lab report. Start from a ready-made type if one fits.
2
Confirm the suggested fields
DocWise proposes the fields, types and validation rules it sees in your samples. Rename, add or remove fields, then test against your samples before you publish.
3
Automate the rest
Every new document is classified, extracted and validated. Confident results flow to your systems by API, webhook or export; uncertain ones wait for a quick human review.

What you get

Classification across your types

Mixed inboxes are sorted into the document types you enable. Documents that match nothing — or several things — are flagged instead of guessed.

Typed values with evidence

Dates, amounts, enums and line-item tables come back as typed JSON. Each value carries a confidence score and the page and text it was read from.

Validation you define

Required fields, formats, ranges and cross-field checks such as line items summing to the total run on every result before it is delivered.

Human review where it matters

Results below your confidence threshold go to a review queue. Reviewers see the source next to the fields and fix only what needs fixing.

Your choice of AI

Extraction runs on the AI provider configured for your workspace — OpenAI or Anthropic models today — with fallback between providers when one is unavailable.

Built to integrate

A documented REST API, signed webhooks with replay, and CSV, JSON or Excel export per document type with line items flattened.

Your documents stay yours

Customer documents are not used to train AI models. AI providers are called with no-training settings where they offer them. By default source files are deleted after 30 days and results after one year; both periods are configurable per workspace.

How we handle data

Pay for pages, not seats

Plans include a monthly page allowance with a clear overage price, so cost follows the documents you actually process. Start free, upgrade when you need more.

See pricing

Try it on your own documents

Create a workspace, upload a few samples and see the fields DocWise suggests.

Start free