AI receipt scanning for Arabic and English bills in the UAE

For each receipt or bill, HelloBooks’ AI looks for the fields an accountant needs to post it and claim VAT correctly. It reads them from a phone photo, a scan or a PDF, in Arabic, English or a mix of the two.

• Supplier name, in the script it appears in.
• Supplier TRN, the 15-digit tax registration number.
• Invoice or receipt number and date.
• Net amount, VAT amount and total in AED, or in the invoice currency.
• Line items where the document lists them clearly.

Extraction is only half the job. Once the fields are read, the AI proposes an account and a VAT treatment based on the supplier, the items and how this business has recorded similar bills before. A bill from a building materials supplier might be proposed as cost of sales with 5% recoverable VAT; a parking receipt as motor expenses.

The image stays attached to the transaction. When an auditor, the FTA or the owner asks for support for a AED 7,800 expense, the document is one click away rather than in a box of paper.

Arabic documents raise particular issues. Numbers may appear in Eastern Arabic numerals, dates may follow different formats and supplier names may be written differently from how they appear on the bank statement. A trader called مؤسسة النور للتجارة on the invoice might appear as AL NOOR TRDG EST on the bank line.

The AI handles numerals and common layouts and learns the link between Arabic supplier names and their bank narrations as the bookkeeper confirms matches. Faded thermal receipts and handwriting remain hard for any system; those are flagged for a person to check rather than posted with low confidence.

Input VAT recovery depends on holding valid tax invoices, and the supplier’s TRN is one of the details a tax invoice must show. A scanner that reads the amount but misses the TRN creates a problem at VAT time.

The AI extracts the TRN where present and flags bills where it is missing or unreadable. Before the quarter closes, a bookkeeper can ask “list bills this quarter where input VAT was claimed but no supplier TRN was captured” and fix them while the supplier can still issue a corrected invoice.

A contracting company in Ajman receives 220 bills a month: 90 in Arabic, 70 bilingual and 60 in English. Manually, a bookkeeper fluent in both languages might spend three minutes per bill, about eleven hours a month, and the firm needs an Arabic reader on every client like this.

With AI scanning, the bookkeeper reviews extracted fields and proposals at perhaps 40 seconds per bill, about two and a half hours. Twenty bills are flagged for closer checks, mostly handwritten delivery notes. The time saved is roughly eight hours a month for this one client, or about AED 10,000 a year at an internal cost of AED 105 an hour.

Photo quality matters more than anything else. Ask clients to photograph receipts flat, in good light, with all four corners visible, on the day they receive them. Thermal receipts fade within weeks, so same-day capture is not just tidier but necessary.

Forward supplier PDFs rather than printing and scanning them. A native PDF is read more accurately than a photo of a printout. If your firm wants specific fields captured for a client, such as a project code written on every bill, ask the HelloBooks team at hello@hellobooks.ai.

Review the extracted totals against the image, especially for large amounts. Check the VAT treatment where the supplier is new or the items are unusual. Confirm that the document is a tax invoice, not a quotation or delivery note, before input VAT is claimed.

Our view: AI scanning removes the language barrier and the keying, but not the need for a trained eye on the result. It turns a bookkeeper from a typist into a reviewer, which is a better use of their time.