Phone: 0850 346 41 09 | Weekdays 10:00–18:00 · Technical support 24/7
Document OCR reads the text on a scanned or PDF document and splits it into fields such as date, amount, tax number and line items. The system we set up collects invoices, delivery notes, contracts and forms from email, a folder or a scanner and posts the fields to your ERP or accounting software. Uncertain fields go to a person for approval.
Your accounts team keys in supplier invoices, the warehouse types up delivery notes, sales re-enters application forms. As volumes grow, so does the time spent, and a mistyped amount or tax number resurfaces at month-end as a reconciliation problem. Tracing which figure came from which document, and who entered it, gets harder too.
We set up a pipeline that collects documents from email, a watched folder or a scanner, reads the text with OCR, uses AI to identify fields such as dates, amounts, tax numbers and line items, checks them, and posts them to your ERP or accounting software. Every record stays linked to its source document. Note that Turkish e-invoices (e-Fatura, e-Arşiv) arrive as structured XML and need no OCR at all; we read that data directly and feed it into the same flow.
Supplier, date, amount, tax number and line items are read from paper or PDF invoices. If totals do not add up or the supplier is unknown, the record is held for approval instead of posted.
Product codes, quantities and batch numbers are read and matched against the purchase order, so short or over-deliveries show up at goods-in. Barcodes and QR codes provide a cross-check.
Parties, value, start and end dates, renewal and termination terms are extracted. Date-driven obligations become reminders in your calendar or CRM.
Fields from ID cards, passports, driving licences and application forms are captured, cutting manual entry in customer onboarding, insurance applications and memberships.
1D/2D barcodes, QR and DataMatrix codes are read and serial number, lot, production date and SKU are recorded. On production lines it works alongside our computer vision systems.
Structured data is extracted from reports, lab results and prescriptions and passed to the hospital information system. Health data is special-category personal data under Turkish law (KVKK), so access is restricted by role.
Documents are picked up automatically from email attachments, a watched folder, a scanner or a web upload.
Skewed, shadowed or low-contrast pages are corrected before reading.
The OCR engine extracts the text; AI identifies the document type and fields such as date, amount and tax number.
Low-confidence fields go to a person for approval. Approved data is written to your ERP or accounting software through API integration.
A clean PDF and a crumpled photo of a delivery note will never give the same result, so we do not promise a blanket accuracy figure. In the pilot we measure field-level accuracy on your real documents and review the results with you. Uncertain fields are routed for approval, so bad data does not slip silently into your system. For non-standard forms we can train a model on your own documents. If your invoices arrive by email, SmartMail in the Smart360 suite already extracts parties, dates, VAT and totals from invoice PDF attachments without downloading the file; you can buy SmartMail through Smart360.
If your documents already end up in a shared archive, SmartFiles enterprise file management analyses every uploaded document automatically: it pulls invoice number, date, seller, buyer, tax ID, tax base, VAT and total into a single table you can export as CSV, reads e-Invoice and e-Archive XML files directly, and handles scanned PDFs and receipt photos with image analysis. To start straight away with a ready-made product, you can buy SmartFiles through Smart360; if the extracted data needs to reach your ERP through an approval flow, we build the custom set-up described on this page.
Supplier invoices reach the accounts team as a mix of paper, PDFs and email attachments, while e-invoices are downloaded separately from a portal. In the pipeline we set up, every document comes in through one intake: e-invoice XML is read directly, PDFs and scans go through OCR, and any invoice whose totals do not add up or whose supplier is unknown lands on an approval list. Instead of keying in every line, the team only looks at the flagged fields. Because each record posted to the ERP stays linked to its source document, at month-end you can trace any figure back to the document it came from.
For more detail, read our OCR invoice processing guide, our article on e-invoice integration in Turkey and our guide to AI document analysis. All related articles are on the AI guides page.
Share a few sample documents; we will show you how the fields are read on your own paperwork and send the scope in a written proposal.
Document OCR reads the text on a scanned or PDF document and splits it into fields such as date, amount, tax number and line items. The system we set up collects invoices, delivery notes, contracts and forms from email, a folder or a scanner, uses AI to recognise the fields and posts them to your ERP or accounting software. Uncertain fields go to a person for approval.
Incoming invoices and receipts, dispatch and shipping documents, contracts and legal documents, ID cards, passports, driving licences and application forms, barcode, QR and DataMatrix labels, and medical records such as reports, lab results and prescriptions. Health data is special-category personal data under Turkish law (KVKK), so access to those documents is restricted by role.
Only after it is checked. If totals do not add up, the supplier is unknown or a field has a low confidence score, the record is held for human approval instead of being posted. Approved data is written to your ERP or accounting software via API, and every record stays linked to its source document so you can trace where each figure came from.
No. e-Fatura and e-Arşiv invoices arrive as structured XML, so no OCR is needed. We read that XML data directly and feed it into the same flow as your paper and PDF documents. OCR is used for documents that arrive on paper, as scans or as PDFs.
A clean PDF and a crumpled photo of a delivery note never give the same result, so we do not promise a blanket accuracy figure. Skewed, shadowed or low-contrast pages are corrected before reading. In the pilot we measure field-level accuracy on your real documents and review the results with you. For non-standard forms we can train a model on your own documents.
Not always. SmartMail, part of the Smart360 suite, extracts the parties, issue and due dates, subtotal, VAT, total, document number and tax number from invoice PDF attachments without downloading the file. For paper, scanned or other documents that must be posted to your ERP, we set up an OCR pipeline.
We look at your document types, monthly volume, the channels documents arrive through and the ERP or accounting software the data should reach. Once accuracy has been measured on samples of your own documents, you receive a written proposal with scope, stages and price. If archiving invoices and getting their key fields as a table is enough, without posting to an ERP, ready-made SmartFiles is also an option.
Have a different question? Ask Us