AI Data Extraction
Shaari uses AI to read your uploaded invoices and automatically fill in the invoice fields. This eliminates most manual data entry and reduces errors from mistyped numbers or dates.
How It Works
When you upload an invoice file (PDF or image), Shaari sends the document to its AI extraction engine. The AI analyzes the document's layout, text, and structure to identify and extract key invoice data.
The extraction happens automatically, with no action required from you beyond uploading the file. Results typically appear within a few seconds.
Fields Extracted
The AI attempts to extract the following fields from each invoice:
| Field | Description |
|---|---|
| Supplier name | The company or individual that issued the invoice |
| Invoice number | The supplier's invoice reference number |
| Invoice date | The date the invoice was issued |
| Due date | The payment due date, if present |
| Subtotal | The amount before tax |
| Tax amount | VAT or other tax amounts |
| Total amount | The final total including tax |
| Line items | Individual products or services with descriptions, quantities, unit prices, and amounts |
| Currency | The currency mentioned on the invoice |
| VAT number | The supplier's VAT registration number |
Not every invoice contains all of these fields. The AI extracts what it can find. Missing fields are left blank for you to fill in manually if needed.
Re-Analysis
If you update an invoice's attachments or want to try extraction again with improved AI suggestions, you can trigger a re-analysis:
- Open the invoice detail view.
- Click the Re-analyze button (or the AI analysis action).
- The AI processes the document again and updates the extracted fields.
Re-analysis is useful when:
- You uploaded a clearer version of the same invoice.
- The initial extraction missed some data.
- You want to compare the AI's suggestions against what you entered manually.
Re-analysis overwrites the AI-suggested values but does not discard any manual edits you have already confirmed. You always have the final say on what data is saved.
Accuracy and Manual Corrections
The AI extraction is designed to be highly accurate, but no automated system is perfect. Common situations where manual correction may be needed:
- Handwritten invoices -- The AI works best with printed or typed text.
- Poor image quality -- Blurry, dark, or skewed images reduce accuracy.
- Unusual layouts -- Invoices with non-standard formatting may confuse the extraction.
- Mixed languages in a single field -- While the AI handles Arabic and English well, mixing both in one field can occasionally cause issues.
After extraction, always review the populated fields before confirming the invoice. Corrections can be made directly in the invoice form.
Supported Languages
Shaari's AI extraction supports:
- Arabic -- Full support for Arabic-language invoices, including right-to-left text.
- English -- Full support for English-language invoices.
Invoices that contain both Arabic and English text (which is common in Saudi Arabia) are handled correctly. The AI identifies each language in context and extracts data accordingly.
For the best extraction results, upload high-resolution, well-lit images or native PDF files rather than scanned copies. The clearer the source, the more accurate the extraction.