Document AI

Invoice OCR
- reinvented

AI-native extraction. No templates, no setup. Drop in a PDF and get structured invoice data back — with conversational feedback to tune it in your natural language.

Price down to

€0,025

/page

API-first
Custom fields, no training
EU/US regions
VertexVision Inc.
Kobenhavn K · DK
Invoice
#INV-2026-0418
Issued · 12 May 2026
Bill to Acme Corp.
Line itemQtyNet
Vector ETL · Subscription11 280,00
Implementation · 5h5650,00
Premium support · Q21425,00
Net total2 355,00
VAT 25%588,75
Total2 943,75
K
Kaunt Document AI
grounded · doc #INV-…0418
Enterprise Flexibility

Tell it what to extract

Built for enterprise requirements. Adjust output with a chat-based interface to serve your customers' needs.

Setup

0 templates

No vendor profiles needed. No mapping wizards. Send a PDF, get structured JSON back.

Custom data

Unlimited custom fields

Add any field, header or line-level — no retraining, no per-field cost.

—Smarter extraction

Smarter Extraction.
Lower cost.

Enterprise-ready premium invoice OCR built for the agentic AI era. Superior accuracy out-of-the-box, no templates, and instant conversational feedback - at unbeatable pricing.

  • AI-native pipeline. Built on AI-powered OCR-engines and fine-tuned LLMs, not 90s template engines.
  • Header & line-level. Structured JSON with every standard invoice field.
  • Custom fields. Go beyond standard invoice fields with prompt-based extraction.
VertexVision Inc.
Kobenhavn K · DK
Invoice
#INV-2026-0418
12 May 2026
Bill to Acme Corp.
Line itemQtyNet
Vector ETL · Subscription 1 1 280,00
Implementation · 5h5650,00
Premium support · Q21425,00
Net total2 355,00
VAT 25%588,75
Total2 943,75
—Built for the real world

Even the trickiest scans.

Multi-vehicle service invoices. Hourly labor with computed line totals. Two address blocks. Subtotals parked in a separate footer table. The kinds of invoices that broke your last OCR. Kaunt extracts them out of the box — no template, no setup, no complaints from your AP team.

St. George Auto Mechanics invoice — 3 vehicles, 14 line items, footer totals
3 address blocks
Sub-grouped by vehicle
Hourly labor, computed
Totals in footer table
Real customer invoice · 3 vehicles, 14 lines, hourly labor
extracted.json
vendorSt. George Auto Mechanics
buyerACME Corporation
invoice.no4234232
remit-to.iban021000021 · JPMorgan Chase
vehicles[0] Toyota Camry · DLP-1234
lines[0]Oil & filter change · $89.00
lines[1]Brake pad replacement (front) · $250.00
labor2.5h × $95.00 = $237.50
vehicles[1] Ford F-150 · CBB-8765
lines[0]Transmission fluid service · $180.00
labor3.0h × $95.00 = $285.00
vehicles[2] Honda Accord · ABC-6555
lines[0]30,000-mile service pkg. · $400.00
labor2.0h × $95.00 = $190.00
totals.net$2 730.50
totals.tax$243.01 · 8.9%
totals.grand$2 973.51
1.9s · 14 line items across 3 vehicles · no human in the loop
—Unmatched Flexibility

Tweak extraction
in any plain language.

Enterprises demand customizable solutions. Kaunt Document AI offers full customization without a single click. Use AI-based textual feedback to directly tweak the OCR output to match your requirements — per company, supplier, or buyer.

  • Prompt-based tweaking at company, supplier, or buyer scope.
  • Unlimited custom fields in header and line-level.
  • Invoice-specific LLM layer that knows what an invoice is.
Example A
“Translate all Japanese invoice lines into English.”
scope: Vendor = Panda Corp.
Example B
“For this supplier, split invoice lines that mention ‘mortgage’ into separate items.”
scope: Vendor = MortgagePro Ltd.
Example C
“Expand abbreviation ‘Ltd’ to ‘Limited’ in vendor names."
scope: Company = global

Every invoice field, structured.

Custom fields can be normalized and validated according to your own instructions - they can even be classifications or interpretations of the invoice data.

Header fields include

Document date

Due date

Currency code

Amount excl. VAT

Amount incl. VAT

VAT amount

VAT rate

Invoice number

Order number

Project number

PO number

Contract number

Customer number

Buyer name

Buyer TAX ID

Buyer email

Buyer phone number

Buyer address

Vendor name

Vendor TAX ID

Vendor email

Vendor phone number

Vendor address

Delivery address

Payment method

Bank SWIFT code

Bank IBAN code

Allowances and charges

Line-level fields include

Line number

Item description

Item code

Quantity

Unit of measure

Unit price

Line amount excl. VAT

Line amount incl. VAT

Line VAT amount

Line VAT rate

Custom Field examples

Service period

Delivery terms

Approval reference

Project code

Buyer department

Employee ID

Service location

Vehicle number plate

VertexVision
#INV-2026-0418
Line itemNet
Vector ETL · Subscription1 280,00
Implementation · 5h650,00
Premium support · Q2425,00
Posting #10.94
GL account4202
Tax code0425
Cost center200 · Engineering
Amount€1 280,00
Posting #20.91
GL account6580
Tax code0425
Cost center300 · IT Ops
Amount€650,00
Posting #30.89
GL account6585
Tax code0425
Cost center300 · IT Ops
Amount€425,00
—Predict account dimensions

Coded by the time
it lands.

Document AI plugs straight into our best-in-class AI Coding service. Predict GL accounts, cost centers, tax codes, and any custom dimension — alongside extraction, in one call.

  • Per-line posting with confidence, splits, and recommended posting text.
  • Tenant-specific learning - trains on your chart of accounts.
  • Hands-off touchless - structured invoice + suggested coding in one response.
See Coding AI
—Chat with your Document AI

Conversational. Grounded. Live.

Drop a document. Describe how extraction should change — in plain English. The agent proposes a rule, shows you the diff, and applies it on confirm. No prompts to engineer, no glue code to write.

invoice_INV-2026-0418.pdf
FieldValue
vendor.nameVertexVision Inc.
vendor.taxIdDK-32194875
invoice.noINV-2026-0418
invoice.date2026-05-12
totals.net2 355,00
totals.vat588,75
totals.grand2 943,75
lines[0].descVector ETL · Subscription
lines[1].descImplementation · 5h
lines[2].descPremium support · Q2
{
  "vendor": {
    "name": "VertexVision Inc.",
    "taxId": "DK-32194875"
  },
  "invoice": {
    "no": "INV-2026-0418",
    "date": "2026-05-12"
  },
  "totals": {
    "net": 2355.00,
    "vat": 588.75,
    "grand": 2943.75
  },
  "lines": [
    { "desc": "Vector ETL · Subscription", "qty": 1, "net": 1280.00 },
    { "desc": "Implementation · 5h", "qty": 5, "net": 650.00 },
    { "desc": "Premium support · Q2", "qty": 1, "net": 425.00 }
  ]
}
updates as rules are applied
K
Document AI agent
grounded on this document
—Why Kaunt

Built for tomorrow.
Not for yesterday.

Legacy OCR was templates — brittle, expensive, painful to configure. Kaunt Document AI is the AI-native rebuild — lower cost, higher accuracy, and tweakable in plain English.

Kaunt Document AI

  • AI-AI-native, template-free
  • Unmatched pricing — down to €0,025 / page
  • Superior accuracy out-of-the-box
  • Flexible, conversational customization
  • Custom fields on header & line-level
  • Text-based feedback API
  • API-first — with an embeddable UI when you want it

Legacy OCR

  • Template-based setups, vendor by vendor
  • High, outdated cost structures (€0,10 – €0,50 / page)
  • Inconsistent accuracy across suppliers
  • Rigid structures, time-consuming configs
  • No custom fields
  • No feedback loop — reach for the consultant
  • Bolted-on UI you have to learn separately
—Built API-first

A clean REST-API. Nothing else required.

Kaunt Document AI is designed for seamless integration. Send a PDF or image, get back structured JSON in seconds. No templates, no training, no extra platform — just one stateless REST call that scales across tenants, suppliers, and regions.

  • Stateless — call it from your ERP, finance system, or workflow.
  • EU & US data residency — data stays where you are.
  • Embeddable UI —want a UI? Drop in the Document AI component.
document-ai · live trace idle
PDF
invoice_INV-2026-0418.pdf
3 pages · 412 KB
POST /v1/document-ai/extract
200 OK // 1.4s · €0,025 { "vendor": { "name": "VertexVision Inc." }, "invoice": { "no": "INV-2026-0418", "date": "2026-05-12" }, "totals": { "grand": 2943.75 }, "lines": [ /* 3 items */ ], "confidence": 0.97 }
OK structured invoice — ready for your ERP
—Embeddable UI

Want the UI? Embed it.

The Document AI Feedback Agent you just saw is a JWT-scoped UI component. Mint a session in your backend, render an iframe, ship the feature.

🔒partner-erp.com/invoices/INV-2026-0418
Partner ERP
Invoices · INV-2026-0418
v4.2 · M
Invoice details
VendorVertexVision Inc.
Vendor IDDK-32194875
BuyerAcme Corp.
NumberINV-2026-0418
Date12 May 2026
Due date11 Jun 2026
CurrencyEUR
Net€2 355,00
VAT 25%€588,75
Total€2 943,75
Lines3 items
Ready to post
Applied rules
Vendor-scope · VertexVision Inc.
Currency · force EUR
+ projectCode (pending)
kaunt · feedback-agent
K
Kaunt Document AI
grounded on this document
Tell me what should change about how this invoice is extracted — I’ll propose a rule you can preview before applying.
Add a custom field projectCode from the PO number.
Done — projectCode added as a header-level custom field. Source: PO number, first 3 chars. Scoped to VertexVision Inc.
Tweak extraction in plain English…
Mint a session · render <iframe> · ship the feature. Same handshake for every Kaunt component.

Invoice OCR hasn’t kept up with technology. The products with templates are outdated, and the ones without templates have never worked well enough. We’ve re-imagined invoice OCR for 2025: AI-native, ERP-ready, and built for the future of automation.

Mads Ellersgaard Kalør, CTO @ Kaunt

Experience OCR re-imagined. AI-native and built for the future of automation.