ESAP
Powered by ESAP AI

From Messy Arabic Documents to Intelligent Data

Upload invoices, quotations, and contracts in Arabic. Our engine extracts, structures, and indexes them for instant retrieval and cited answers.

ESAP · Chat
R
Does invoice INV-2024-0341 match PO 2024-118 and delivery note DN-0341?
Yes — invoice total SAR 42,850.00 INV-2024-0341 · p.3 matches the purchase order amount PO-2024-118 · p.1 and goods received on 12 Mar 2024 DN-0341 · p.1.
Ask about your documents…
INVOICEINV-2024-0341 · p.3
TOTALSAR 42,850.00
PURCHASE ORDERPO-2024-118 · p.1
AMOUNTSAR 42,850.00
DELIVERY NOTEDN-0341 · p.1
RECEIVED12 Mar 2024

Trusted by teams across the region

Aramco Digital
STC Cloud
Majid Al Futtaim
Riyadh Municipality
Mashreq Bank
Emirates NBD
The Challenge

Real-world Arabic documents are chaos

  1. Handwritten Invoices

    Scribbled numbers and Arabic calligraphy that OCR can't parse.

  2. WhatsApp Photos

    Blurry phone photos of receipts sent in chat messages.

  3. Scanned PDFs

    Low-resolution scans with stamps, watermarks, and noise.

  4. Mixed AR / EN

    Documents mixing Arabic and English with inconsistent direction.

  5. Inconsistent Formats

    Every supplier uses different layouts, tables, and structures.

  6. Multi-page Contracts

    Long documents spanning pages with tables, terms, and signatures.

How It Works

From chaos to clarity in 4 steps

01
Upload

Drop any Arabic document — PDF, photo, scan. We handle the chaos.

02
Extract

ESAP's extraction engine reads every character, stamp, and annotation with human-level accuracy.

03
Structure

Validated JSON: issuer, recipient, line items, totals, dates — bilingual and normalized.

04
Query

Ask in Arabic or English — get cited answers from your document corpus instantly.

Built for Arabic document intelligence

Intelligent Extraction

Our extraction engine reads documents like a human — handwriting, stamps, logos, tables.

Deep Analysis

Multi-pass reasoning for Arabic character disambiguation — ب vs ث, ت vs ش, correctly every time.

Hybrid Search

Multi-signal retrieval combining semantic understanding and keyword matching for maximum recall.

Bilingual

Ask in Arabic or English, search across both. Answers reference the original language.

Structured Output

Clean JSON for every document: parties, line items, totals, dates, terms, and notes.

Live Processing

Upload any document and watch it get extracted, structured, and indexed with live progress.

Bilingual parity

Arabic is not a fallback.

Both scripts render at equal weight in every context.

Invoice INV-2024-0341

LTR
Issuer
United Transport Co.
Date
15 Mar 2024
Total
SAR 42,850.00

فاتورة رقم ٠٣٤١

RTL
الجهة المُصدِرة
شركة النقل المتحد
التاريخ
١٥ مارس ٢٠٢٤
الإجمالي
SAR 42,850.00
The numbers

Proven on real Arabic enterprise docs.

0+Documents processed
0%Bilingual extraction accuracy
0msMedian query latency

Stop re-typing invoices.

Start asking them questions instead.

Try PageSense
Talk to sales