Field-level extraction
Cheque number, date, payee (Arabic & English), amount in numbers and words, and account number.
Cheque Intelligence · Banking Document AI · Powered by AIVortex
Cheque Intelligence — banking documents, read and validated.
Cheque Intelligence extracts cheque fields on-prem — Arabic and English OCR with template and ROI detection, a confidence score on every field, and a review-and-retrain loop for the cases the model isn't sure about. Not a generic OCR demo: built for bank cheque formats.
Cheque OCRField-level extractionConfidence per fieldArabic & EnglishReview & retrain loopOn-prem
Cheque Intelligence · Banking Document AI
ExtractValidateFieldsAudit
Batch · Operations
“Extract the fields from this corporate cheque and flag anything below our 0.90 threshold.”
Cheque Intelligence · field-level extract
Cheque no. 012345 · date 2026-01-29 · payee ABC Trading Co. (شركة ايه بي سي للتجارة) · amount 1,250.00. Six of seven fields cleared the threshold; amount-in-words at 0.89 was routed to the review queue.
cheque_number · 0.98amount_numeric · 0.96amount_words · 0.89
Route to Verification Clerk Review
audit · logged · 09:18:44 · template=CORP-A · flagged=1
Capabilities
Built for the realities of banking documents.
Cheques come in many templates and two scripts. Cheque Intelligence reads them on-prem, extracts the exact fields banking operations need, scores its confidence on each one, and gets better over time from the corrections your team makes.
Confidence per field
Every value carries a confidence score — set a threshold for straight-through processing versus review.
Template & ROI aware
Region-based extraction tuned to your cheque layouts — including Qatar-region and bilingual templates.
Human review queue
Items below your threshold route to validation with the original image — and corrections are captured.
Continuous improvement
Captured corrections become labeled data for retraining — measurable accuracy uplift on your hard cases.
API-first & on-prem
Structured JSON and CSV via REST to core banking, ECM/DMS and workflow engines — running fully inside your network.
How it works
From scanned cheque to validated data.
01 — Ingest
Upload or batch
Process scanned images or PDFs from your capture systems, with optional pre-processing.
02 — Detect
Template & ROI
Identify the cheque format and locate the regions for each required field.
03 — Extract
OCR + post-process
Arabic/English OCR with cleaning, validation rules and a confidence score per field.
04 — Decide
Auto-approve or review
High confidence goes straight through; low confidence routes to review and captures corrections.
You set the threshold. Straight-through where the model is confident, human review where it isn't — and every correction feeds the improvement loop. Nothing low-certainty is passed downstream unchecked.
Built for
For banks and back offices.
Banking operations
Automate cheque and document data entry across branches and processing centres.
Trade & payments
Extract and validate data from payment instruments and supporting documents.
Shared services
Cut manual keying and errors across high-volume document workflows.
Part of the same document stack as the Document Intelligence Hub.
Let's build
See Cheque Intelligence on your documents.
Bring a batch of real cheques or banking forms and we'll show on-prem extraction, confidence scoring and validation on your own templates.